Draft version January 3, 2018 Preprint typeset using LATEX style emulateapj v. 12/16/11

HABITABILITY OF EXOPLANET WATERWORLDS Edwin S. Kite Department of Geophysical Sciences, University of Chicago, Chicago, IL 60637, USA [email protected]

Eric B. Ford Department of Astronomy and Astrophysics, The Pennsylvania State University, University Park, PA 16802, USA Center for Exoplanets and Habitable Worlds, The Pennsylvania State University, University Park, PA 16802, USA Institute for CyberScience and Pennsylvania State Astrobiology Research Center Draft version January 3, 2018

ABSTRACT We model the evolution of ocean temperature and chemistry for rocky exoplanets with 10-1000× Earth’s H2O but without H2, taking into account C partitioning, high-pressure ice phases, and atmosphere-lithosphere exchange. Within our model, for Sunlike stars, we find that: (1) habitability is strongly affected by ocean chemistry; (2) possible ocean pH spans a wide range; (3) exsolution-driven climate instabilities are possible; (4) surprisingly, many waterworlds stay habitable for >1 Gyr, and (contrary to previous claims) this longevity does not necessarily involve geochemical cycling. We also find, using an ensemble of N-body simulations that include volatile loss during giant impacts, that a substantial fraction of habitable-zone rocky planets emerge after the giant impact era with deep, ice-free water envelopes. This outcome is sensitive to our assumptions of low initial 26 60 abundances of Al and Fe in protoplanetary disks, plus H2-free accretion. We use the output of the N-body simulations as input to our waterworld evolution code. Thus, for the first time in an an end- to-end calculation, we show that chance variation of initial conditions, with no need for geochemical cycling, can yield multi-Gyr habitability on waterworlds. Keywords: planets and satellites: terrestrial planets — planets and satellites: physical evolution — planets and satellites: individual (Kepler-452b,Kepler-1638b, Kepler-1606b, Kepler-1090b, Kepler-22b, τ Ceti e, Proxima Cen b, TRAPPIST-1, GJ 667 C, LHS 1140 b, Ross 128 b, Kepler-62f, Kepler-186f, GJ 832 c, HD 40307 g, Kepler-442b, Kepler-1229 b)

1. INTRODUCTION. (Foley 2015; Abbot et al. 2012). Yet waterworlds should Habitable-zone small-radius exoplanets are common be common in the Galaxy (e.g. Mulders et al. 2015). Are (Burke et al. 2015; Dressing & Charbonneau 2015). waterworlds doomed? Not necessarily, because there is What fraction are habitable? For the purposes of this another (less-studied) road to long-term planetary hab- paper, an operational definition of a potentially habit- itability: able exoplanet is: maintains T < 450K liquid water on Cycle-independent planetary habitability: Planets its surface continuously for timescales that are relevant 7 where pCO2, ocean depth, and surface temperature Tsurf for biological macroevolution, 10 yr (Vermeij 2006; 7 1 remain within the habitable range for 10 yr without Carter 1983; Bains et al. 2015). With this definition, geochemical cycling. two pathways allow long-term planetary habitability: Habitability maintained by geochemical cy- In this paper, we first show that the long-term climate cles: Planets that maintain habitability despite large- evolution of waterworlds can be modeled independently of geochemical cycles (§2). Next, we set up (§3) and

arXiv:1801.00748v1 [astro-ph.EP] 2 Jan 2018 amplitude perturbations (due to tectonics, stellar evolu- tion, e.t.c.) through a negative feedback. A proposed run (§4) such a waterworld evolution model. Using the mechanism for the negative feedback is a geochemical model, we estimate what combinations of water abun- cycle involving volcanic outgassing and the fixation of dance, initial carbon abundance, and geologic processes atmospheric gases into rocks (carbonate-silicate weath- allow habitable surface water to persist for >1 Gyr. We ering feedback; Walker et al. 1981; Kasting et al. 1993). focus on CO2+H2O(±N2) atmospheres (Wordsworth & Pierrehumbert 2013). We ignore H2 warming (Stevenson Climate-stabilizing geochemical cycles are a mainstay 1999). We emphasize long-term climate evolution, and of textbooks and review papers (Catling & Kasting defer discussion of nutrients to §6. In §5, we use an N- 2017; Knoll et al. 2012; Kaltenegger 2017, and refer- body model of planet assembly to demonstrate how wa- ences therein). However, such feedbacks probably do not terworlds form and migrate to the Habitable Zone (HZ). work on habitable-zone rocky planets with water mass- We discuss in §6 and conclude in §7. Figs. 15–17 show fractions 10× - 103× that of the Earth – “waterworlds” our main results.

1 Sub-ice oceans in extrasolar planetary systems may be habit- able, but this cannot be confirmed from Earth by remote sensing. 2 Kite & Ford

1.1. This paper in context Simulations of rocky-planet formation yield many rocky small-radius (Rpl <1.6 R⊕) planets with water 2 mass fractions fW = 10-1000 × fW,⊕ (Raymond et al. 2004, 2007; Ciesla et al. 2015; Mulders et al. 2015). −4 Earth’s fW (∼ 10 ) appears to be the result of chance (e.g., Raymond et al. 2007; Morbidelli et al. 2016; Licht- enberg et al. 2016) (§5.2). Indeed, simulations suggest fW ≈ fW,⊕ may be uncommon on planets in general (e.g. Tian & Ida 2015). Among previous studies of waterworlds (e.g., Kuchner 2003; L´egeret al. 2004; Selsis et al. 2007; Fu et al. 2010; Abbot et al. 2012; Levi et al. 2017), the closest in in- tent to our own are those of Kitzmann et al. (2015) and Noack et al. (2016). Kitzmann et al. (2015) consider carbonate-system equilibria and find a “sweet-spot” in total planet C similar to our own, but they do not con- sider geological processes. Noack et al. (2016) empha- size the development of high-pressure H2O-ice (HP-ice) layers that isolate the liquid ocean from the nutrients supplied by silicate-rock leaching. Noack et al. (2016) also confirm the result of Kite et al. (2009) that deep oceans suppress silicate volcanism. We conservatively do not count planets with HP-ice as examples of habitability Figure 1. Sketch showing response of an Earth-mass planet to (see §6.4 for discussion). For Tsurf . 375K, this sets a perturbations of the geologic C cycle for (upper panel) an planet Tsurf -dependent upper limit on the depth of a habitable with shallow oceans but no geochemical feedbacks versus (lower ocean (Fig. 1.1), and we consider only <8 GPa oceans. panels) a waterworld. The two temperature-versus-time tracks for We go beyond Noack et al. (2016) by considering the each panel correspond to an increase (upper track) and a decrease effect of C on climate, by using an N-body code to cal- (lower track) in the CO2 flux from the lithosphere into the atmo- sphere+ocean. Color shows fraction of initial H2O lost to space, culate how volatile delivery and giant impacts “load the assuming XUV-energy-limited escape Lammer et al. (2009) when dice” by regulating the fraction of planets that can have the dimensionless moist number >1 (Wordsworth & Pierrehum- cycle-independent planetary habitability, and – most im- bert 2013). In the shallow-ocean case, the climate rapidly becomes uninhabitable (so in this toy model geochemical feedbacks are re- portantly – by tracking Tsurf and pCO2 for 10 Ga on the quired to restore and sustain habitability). In the deep-ocean case, habitable worlds we model. the planet remains habitable for many Gyr due to the subdued Most of the physical and chemical processes we dis- geologic C-cycle, plus ocean dilution. cuss have been investigated previously in an Earth con- text. The novel aspect of our paper is the application to exoplanet waterworlds (§2).

2. HOW TO MODEL HABITABILITY ON WATERWORLDS.

Waterworlds are buffered against H2O loss (§2.1), but because CO2 also modulates planetary habitability (§2.2), we need to consider the chemistry that sets ocean pH and thus CO2 (§2.3). Fortunately, nature makes this task easier on waterworlds (§2.4), by chemically isolating the ocean+atmosphere from the lithosphere after 108 yr. This isolation permits cycle-independent planetary hab- itability.

2.1. Waterworlds are buffered against H2O loss For an initial water endowment of >50 Earth oceans in the habitable zone (HZ) of a Sunlike star, the moist greenhouse cannot cause complete ocean loss because XUV-limited water loss in the G-star habitable zone can- Figure 2. Cartoon showing the effect of partial pressure of not remove the bulk of the ocean (Kasting 1988; Lammer atmospheric CO2 and insolation L∗ on habitability, based on et al. 2009; Zahnle & Catling 2017) (Fig. 1). The XUV Wordsworth & Pierrehumbert (2013). The planet fates define a limit may be too pessimistic, in part because photolysis lozenge of habitability that is broadest for O(1) bar pCO2. Color can be limiting for high UV fluxes (e.g. Wordsworth ramp corresponds to Tsurf (K), and extends from 273K to 500K. The peak around 8 bars is an interpolation artifact. & Pierrehumbert 2013; Owen & Alvarez 2016. Also,

2 The H mass fraction of the atmosphere+ocean+crust+(silicate mantle), excluding any H in the core, multiplied by 9 to get H2O- equivalent mass fraction. Habitability of exoplanet waterworlds 3

atmosphere-interior exchange of H2O is not important for deep-ocean planets that have solid rock interiors, be- cause the H-storage capacity of solid Earth mantle rocks is only 1-10 Earth oceans (e.g. Hirschmann 2016) – this is swamped by fW on a waterworld (Tikoo & Elkins- Tanton 2017; for a contrary view, see Marty 2012). For such worlds, the ocean you have after cooling from the last giant impact is the ocean you keep.

2.2. Shallow-ocean worlds: vulnerable unless geochemical cycles maintain habitability.

Planetary habitability is modulated by CO2 (Fig. 1). Very high partial pressures of atmospheric CO2 (pCO2) lead to temperatures too high for life, but intermediate pCO2 can stave off glaciation (up to an outer limit set by CO2 condensation and Rayleigh scattering) and thus ex- tend habitability (Fig. 1) (Kopparapu et al. 2013). pCO2 would have varied widely if Earth had always lacked a negative feedback on pCO2 (Kasting & Catling 2003). On Earth, >90% of C is stored in rocks, and C cycles be- Figure 3. How the weight of water inhibits silicate volcanism tween the atmosphere+ocean+biosphere and rocks every on stagnant-lid waterworlds (Kite et al. 2009). The dotted line ∼300 Kyr in the modern era (Knoll et al. 2012) (prob- A→B→C→D tracks the partial (batch) melting of a parcel of ably 100 Myr in the Hadean; Sleep & Zahnle 2001) – initially-solid mantle rock in the rising limb of a solid-state mantle- convection cell beneath a stagnant-lid lithosphere on a no-ocean this is geologically rapid. As a result of this rapid geo- world. From A→B, the path parallels the solid-state adiabat. chemical cycling, a small initial imbalance between the Dashed gray lines show adiabats corresponding to surface tem- rate of C release from rocks (volcanic/metamorphic out- ◦ peratures of 1350 C (modern mantle potential temperature) and gassing) and C uptake into rocks (weathering uptake) 1550◦C (hottest ambient-mantle potential temperature known on Earth, ∼3.0 Gya ; Herzberg et al. 2010). At B, melting begins, and could lead to comprehensive surface glaciation (a snow- increases until (at C) the parcel reaches the base of the 60 km-thick ball) or a moist greenhouse. An initial moist greenhouse stagnant lid (thick gray horizontal line). Further ascent (if any) is with XUV-limited ocean loss would lead to loss of Earth’s slow, and conductive cooling shuts down further melting (at D). ocean water (Fig. 1). Shallow-ocean worlds are vulnera- For a stagnant-lid planet with a 3 GPa ocean, the thin solid line 3 path A→B→C0→D0 tracks a mantle parcel’s ascent. This mantle ble. parcel reaches the base of the stagnant lid (thick blue horizontal Forming with a factor of a few times more water than line) before much melting has occurred. The dashed blue lines the Earth – enough to drown the land, but not as deep are adiabats displaced downwards such that the sub-lithospheric temperature is the same as in the no-ocean case. For details, see as the worlds we model in this paper – makes it more dif- Appendix A. ficult to equalize silicate weathering and CO2 outgassing (Foley 2015; Abbot et al. 2012), and so it is less likely that a small initial imbalance between C release and C uptake could be restored via a negative feedback.

2.3. Calculating waterworld pCO2(t): habitability is strongly affected by ocean chemistry Although waterworlds are buffered against water loss (§2.1), pCO2 is still important (Fig. 1). pCO2 depends on C abundance, fW (≡ dilution), and ocean pH. pH matters because of the strong effect of H+ activity4 on carbonate equilibria (Fig. 5). pCO2 in the atmosphere is in equilibrium with dissolved CO2 in the ocean:

[CO2(aq)] pCO2 = (1) kH

where the square brackets denote concentration, and kH

3 Doom is not assured for a hypothetical shallow-ocean world Figure 4. Maximum ocean pressure to avoid a frozen seafloor without a negative feedback. Neither the moist greenhouse transi- (high-pressure ice, HP-ice). Thick dashed lines correspond to tion nor the snowball transition are completely understood, and it liquid-water adiabats (spaced at 30K intervals in sea-surface tem- is possible that they can self-arrest or reverse (Abbot et al. 2011; perature). Liquid-water adiabats are shown as continuing beyond Abe et al. 2011; Abbot et al. 2012; Hu et al. 2011; Kodama et al. the HP-ice freezing curve (thick solid line). For Tsurf & 375K, the 2015; Hoffman et al. 2017). Also, the desirable pCO2 for a planet liquid-water ocean transitions to supercritical fluid with increasing near the inner edge of the HZ is small. For such a world, the com- P and HP-ice is avoided. As the star brightens, it is possible for bination of vigorous C uptake, feeble C outgassing, and an M-star seafloor HP-ice to melt. Thin dashed line corresponds to the phase host could allow 10 Gyr of habitability. 4 + stability boundary between the calcite (cal) and aragonite (arg) pH ≈ -log10[H ], where the square brackets here denote frac- forms of CaCO3 (Redfern et al. 1989). For details, see Appendix tional concentration. Despite the similar notation, pH has nothing B. to do with the partial pressure of H2. 4 Kite & Ford

evolution independent of ocean-mantle geochemical cy- cling (Fig. 6). The most important factor is that seafloor pressure on waterworlds curtails adiabatic decompression melting, which is the dominant mechanism of volcanism on rocky planets (Kite et al. 2009; Noack et al. 2016) (Fig. 3) (Appendix A). The six factors, which we col- lectively term the “waterworld approximation”, are as follows:

1. H2 is negligible; 2. Water loss to space is small; 3. H2O in the atmosphere+ocean greatly out- weighs H2O species in the silicate mantle; 4. Silicate volcanism shuts down within O(108) Figure 5. Showing C distribution within the aqueous phase com- yr after the last giant impact. puted using CHIM-XPT (Reed 1998) (Bjerrum plot; for 0.1◦C). 5. Deep mantle cycling of C is small; − Solid red line is dissolved CO2, dashed magenta line is HCO3 , and 6. No land. 2− dotted blue line is CO3 . Square brackets denote molalities; Ca and Na are examples of cations. Small wiggles in the curves are We explain each factor below. interpolation artifacts. is a solubility constant. However, the pCO2 can be much 1. H2/He is negligible. H2 blankets of the appropriate less than expected by dividing total inorganic C (i.e., [C]) thickness to enhance habitability in the HZ of Kopparapu in the ocean by kH , if the pH is high enough to form et al. (2013) both require fine-tuning to form, and swiftly HCO− or CO2− at the expense of CO (Fig. 5): escape to space (e.g. Wordsworth 2012; Odert et al. 2017; 3 3 2(aq) but see Ramirez & Kaltenegger 2017). We do not track H2 outgassing by Fe + H2O → FeO + H2. This reac- + 2− + 2− (CO2(aq)+H2O↔H2CO3)↔H +HCO3 ↔2H +CO3 tion requires mixing of Fe with the volatile envelope dur- (2) ing giant impacts, but planetary embryos contain metal When ocean pH is high (H+ concentration is low), then cores of >500 km diameter, which merge quickly (Dahl by Le Chatelier’s principle carbonic acid gives up its H+, & Stevenson 2010; Jacobson et al. 2017). Therefore, we 2− − C is hosted in the ocean as CO3 and HCO3 , and the consider a H2-free, CO2+H2O(±N2) atmosphere. fraction of total C in the atmosphere is small, as on Earth today (Fig. 5; Zeebe & Wolf-Gladrow 2001; Ridgwell & 2. Water loss to space is small. This is explained in Zeebe 2005; Zeebe 2012). In other words, high pH ef- §2.1. fectively sequesters CO from the atmosphere by driv- 2 3. H2O in the atmosphere+ocean outweighs H2O ing the carbonate equilibria to the right. When ocean species in the silicate mantle. The crystallizing pH is low (H+ concentration is high), then by Le Chate- silicate-magma ocean exsolves H2O. The H2O remain- lier’s principle, C in the ocean exists mostly as dissolved ing in the silicate mantle after crystallization is limited CO2 (which includes H2CO3), and the fraction of total by the H O storage capacity of upper mantle minerals, C in the atmosphere is large. 2 2+ + assuming that the mantle convects sufficiently vigorously Ocean pH rises when dissolved rock (e.g, Ca , Na ) that each parcel of mantle passes through the upper man- is added to the ocean. That is because charge balance is tle at least once. This storage capacity is small relative maintained almost exactly in habitable oceans. Addition 2+ + to fW on waterworlds (Hirschmann 2006, but see Marty to an ocean of a mole of Na relieves one mole of H 2012 for a contrary view), so any subsequent cycling (Ko- from their duty of maintaining charge balance, so they + macek & Abbot 2016; Korenaga et al. 2017) will have lit- revert to H2O. Loss of H raises pH, and sucks C out of tle (fractional) effect on ocean depth5. Therefore, after the atmosphere (Eqn. 1, Fig. 5). This effect of dissolved the magma ocean has crystallized, we assume constant rock on pCO2 does not involve the formation of carbonate ocean depth. minerals. However, C can exist mostly as solid CaCO3 if [Ca] is high. 4. Silicate volcanism shuts down within O(108) Assuming constant atmosphere+ocean C content, yr after the last giant impact. To shut down C ocean pH falls when T rises. cycling between a planets’ mantle and ocean, is neces- A general model for pCO2(t) would be complicated sary to shut down volcanic outgassing. To shut down (Holland 1984; Hayes & Waldbauer 2006). pCO2 is cou- volcanic outgassing, it is sufficient to shut down volcan- pled to pH, pH is affected by rock-water reactions, and ism. (Other forms of outgassing such as geysers are ulti- the controls on rock-water reactions are complicated and mately driven by volcanism). Volcanism is curtailed by evolve with time. Fortunately, for waterworlds, the prob- seafloor pressure, were the seafloor pressure threshold is lem is simpler. dependent on the the mode of mantle convection, and is

5 2.4. pCO2(t) modeling is simplified by the waterworld For seafloor pressure > 10 GPa, the H2O storage capacity approximation of mantle rock beneath the stagnant lid jumps to ∼0.5-1.0 wt% (Ohtani et al. 2005), but for such high seafloor pressures the domi- Six factors tend to reduce the number of fluxes that nant H reservoir is still the ocean, because the H2O storage capacity we have to model. Each factor helps to make the climate of average mantle rock is still 0.5 wt%. Habitability of exoplanet waterworlds 5

Figure 6. The flow of our waterworld evolution model. Reservoirs (bold), processes (italics), and variables (regular text) that are discussed in the text are shown. Cycle-independent planetary habitability is enabled by cessation of ocean-mantle geochemical cycling after ∼108 yr on waterworlds.

Table 1 Parameters and variables.

Parameter Description Value Units a semimajor axis AU asl anchor for the evolving snowline 1.6 AU −2 C atmosphere+ocean C in CO2-equivalent kg m cC CO2-equivalent C mass in atmosphere+ocean Fraction of planet mass −1 −1 cp,r specific heat of rock J kg K * −1 [C], [Na], [Ca] aqueous concentrations (molalities) of C; Na; Ca mol (kg H2O) fW ocean mass Fraction of planet mass fW,max maximum fW in planetary embryos Fraction of embryo mass I/E intrusive:extrusive volume ratio (lavas vs. sills/dikes) - −1 −1 kcond thermal conductivity of ocean W m K −2 L∗ stellar luminosity at planet W m Mpl planet mass Earth masses P pressure (we ignore the 1% difference betwen 1 bar and 105 Pa) Pa, or bars Plith pressure at the base of the lithosphere pCO2 partial pressure of atmospheric CO2 bars RP l planet radius km ◦ T , Tsurf temperature; ocean-surface temperature K, or C Tp silicate-mantle potential temperature K t time Gyr W/R effective water/rock ratio of water/rock reactions (see Appendix D) - zcr crustal thickness ∆Z ocean depth km ∆zbl thickness of conductively-cooling layer(s) within ocean km τc, τx, τdyn timescales for steam-ocean cooling; crust formation; and C delivery Gyr τhab duration of habitable surface water Gyr τslurry slurry-layer cooling timescale Gyr −3 ρcr crustal density 3000 kg m −3 ρw ocean density 1300 kg m

* We use mol/kg interchangeably with mol/(kg H2O), which introduces small errors that are acceptable for our purposes. 6 Kite & Ford smaller for stagnant lid convection (∼1-3 GPa) (Fig. 3). 3.1. Setting seawater composition For simplicity, we impose an abrupt 1 GPa cutoff on In this subsection, we provide justification for (1) volcanism. Because volcanism is intimately connected varying atmosphere+ocean CO2 mass fraction, cC , from to plate tectonics, it is plausible that without volcan- 10−5 to 10−3 of planet mass, such that c /f varies6 ism there is little or no tectonic resurfacing (Sleep 2000). C W from 10−6 to 10−1, and (2) using salinities O(0.1-1) Without tectonic resurfacing, the planet’s interior is in mol/kg. We define c to be the mass of C initially in the mode of mantle convection that characterizes Mars, C the atmosphere+ocean (excluding C in the metal core or Mercury, and the Moon: stagnant lid (O’Rourke & Ko- silicate mantle; after the magma ocean has crystallized renaga 2012). In stagnant lid mode, the planet’s internal but without carbonate formation), divided by the mass heat is evened-out by internal solid-state convection, and of the planet, and multiplied by 44/12 to get the CO - released mainly by conduction across a stable, very vis- 2 equivalent mass fraction. Busy readers may skip to the cous boundary layer (the stagnant lid). next subsection. Even without sustained volcanism, water-rock interac- Seawater composition on stagnant-lid waterworlds is tion still occurs on waterworlds. Volcanism will occur 8 set by (1) delivery of CO2, and (2) supply of cations via .10 yr after the last giant impact, e.g. extrusion of 8 the dregs of the magma ocean. This volcanism is key water-rock interaction in the 10 yr after the last giant impact. Volatiles initially in the steam ocean are less to setting ocean chemistry, and is discussed in §3. Slow 7 alteration of porous seafloor rocks, and carbonate pre- important. cipitation/dissolution at the seafloor, can both continue, (1) C delivery. The difference in notation between cC and due to diffusion and/or changes in seafloor T that shift fW is intended to emphasize that for C (unlike H), the mineral equilibria. These slow processes, which we do dominant reservoir can be the metal core. The Fe-metal not model, supply nutrients (§6.4). core takes up C during (liquid metal)-(liquid silicate) equilibration in the immediate aftermath of giant im- 5. Deep-mantle cycling of C is small. After volcanic pacts (Kuramoto & Matsui 1996; Hirschmann 2012; Das- outgassing has shut down, C is not released from the gupta 2013). According to Dasgupta 2013 (his Fig. 2), mantle into the ocean. C is not subducted into the deep C is so iron-loving that, even if 2% of a 1 M⊕ planet’s mantle either, because stagnant lid tectonics is likely. C composition is C (equivalent to 7 wt% CO2), only can still be sequestered by carbonate formation at (or <100 ppm escapes the core. The lure of Fe-metal was just below) the seafloor. As a result, we treat the atmo- so strong that most of the C that existed in the Earth sphere+ocean+(shallow lithosphere) as a closed system prior to the Moon-forming impact is thought to now be with respect to C after 108 yr (Figure 6). trapped in the core. Therefore, the C in our bodies (and in Earth’s ocean) is thought to represent C deliv- 6. No land. Topographic relief is set by the competi- ered after the era of giant impacts (Bergin et al. 2015). tion between tectonic stresses and gravity-driven crustal This is in contrast to the H in Earth’s ocean, which is spreading (Melosh 2011). The tallest peaks in the Solar thought to have arrived during the main stage of planet System are ∼20 km above reference level. Such peaks assembly (O’Brien et al. 2014; Rubie et al. 2015; see Al- are drowned on waterworlds, so there is no land. With bar`ede2009 for an opposing view). Giant impacts are no land, there is no continental weathering. stochastic, so cC /fW is expected to vary stochastically Now, for a given planet mass (Mpl), we can write (Hirschmann 2016) (§6.5): the varying extent of the re- action Fe + H2O → FeO + H2 (which we do not track) is another source of possible variability. Given these uncer- −5 −3 pCO2 =pCO2(fW,t=0.1Gyr, cC,t=0.1Gyr, t, L∗(t), [X]t=0.1Gyr) tainties, we vary cC from 10 to 10 , such that cC /fW −6 −1 −5 (3) varies from 10 to ∼10 . cC = 10 corresponds to −3 where fW is planet water mass fraction, L∗ is stellar lu- an Earth/Venus like C abundance, and we find that 10 minosity, [X]t=0 is initial ocean chemistry, and t is time all but ensures a pCO2 > 40 bar atmosphere, which we in Gyr. Here “t = 0.1 Gyr” really means “t ≤ 0.1 Gyr,” deem uninhabitable. since our simulations have 0.1 Gyr timesteps (Fig. 6). (2) Water-rock interaction: Rock with the composition Neither weathering flux, nor ocean-interior exchange flux of Earth mid-ocean ridge basalt (CaO ≈ MgO ≈ FeO ≈ appear in Eqn. 3, because of the “no land” and “no post- 10 wt%, Gale et al. 2013) can, via carbonate formation, 0.1 Ga volcanism” assumptions, respectively (Fig. 6). potentially sequester up to 70 (bars CO2)/(km crust). To The only time dependence in Eqn. 3 is warming as the what extent is this potential realised? The fluid composi- star brightens.

3. DESCRIPTION OF WATERWORLD EVOLUTION MODEL 6 In this paper we do not consider comet-like, very C-rich com- positions. These require a more sophisticated treatment including We set out to solve Eqn. 3. The overall flow of our clathrates and other C-rich phases (Bollengier et al. 2013; Levi et code is shown in Fig. 6. There are two main stages: al. 2014; Marounina et al. 2017a; Levi et al. 2017). 7 Immediately after the last giant impact, volatiles and magma • Setting seawater composition: random assign- are in equilibrium. Below 1600K, the surface can crust over. ment of “initial” C abundance and cation abun- Volatiles in the atmosphere at this point are likely to be incor- dance from cosmochemically/geophysically reason- porated into the H2O(l) ocean. The most abundant non-CO2 non-H2O volatile in the 100-bar, 1600 K steam-atmosphere cal- able ranges (t <0.1 Gyr) (§3.1), culations of Lupu et al. (2014) is HCl, at volume mixing ratio ≈ 10−2. Upon condensation, if we dilute this column abundance • Calculation of Ts(t) and pCO2(t)(<0.1 - 10 Gyr) into a 100 km-deep ocean (the shallowest we consider) we obtain (§3.2). [Cl] ≈ 10−2 mol/kg. Na is even less abundant. Habitability of exoplanet waterworlds 7 tion resulting from water-rock interaction <108 yr after Na0 = 0.0279 is the Na mass concentration in the crust 3 −3 the last giant impact can be estimated via analogy to (Gale et al. 2013), ρcr = 3 × 10 kg m is crust density, −3 similar worlds, by thermodynamic calculations, and by and ρw ∼ 1300 kg m is ocean density (this is the den- geophysical arguments about the mass of rock available sity of water at 1.8 GPa and 373K; Abramson & Brown for fluid alteration. 2004). For zoc = 100 km, zcr = 50 km, and I/E = 5, this Previous work on global extraterrestrial oceans – such gives [Na] = 0.3 mol/kg. as Europa and Enceladus – suggests total salinities (ii) Pervasiveness: Carbonation of Earth’s modern O(0.1-1) mol/kg. For example, Glein et al. (2015) and oceanic crust is restricted to the vicinity of veins and Hand & Chyba (2007) estimate salinity using constraints cracks (Alt 1995). However, there is evidence from from spacecraft data; and McKinnon & Zolensky (2003), seafloor rocks formed under C-rich, possibly warmer con- Zolotov (2007), and Zolotov & Kargel (2009), estimate ditions ∼3.4 Ga on Earth (Nakamura & Kato 2004) that salinity using models. pervasive alteration of seafloor extrusive rocks can oc- We use the aqueous geochemistry thermodynamics cur. Moreover, Earth’s ocean pH is thought to have been solver CHIM-XPT (Reed 1998) to further explore water- moderate for 4 Ga (7.5±1.5; Halevy & Bachan 2017), and rock interaction (Appendix D). For these runs, we as- either higher pH or lower pH would be more corrosive. sume rock with the composition of Earth mid-ocean- Therefore, a wide range of alteration – including per- ridge basalt (Gale et al. 2013) because basalt is ubiqui- vasive alteration – is possible (Vance et al. 2007, 2016; tous (Taylor & McLennan 2009). CHIM-XPT runs (for Kelemen & Hirth 2012; Neveu et al. 2015). T ≤ 600K) indicate that at low W/R, Mg and Fe form sil- icates and so do not draw down CO . Non-participation Taken together, the above arguments suggest that 2 both Na and Ca, but not Fe and Mg, participate in of {Mg,Fe} cuts C uptake by a factor of ∼3. CaCO3 formation reduces [C] in outlet fluid to 0.1 mol/kg. CO2 drawdown if rocks interact with water; and that . the reasonable range of rock-layer thicknesses for com- Complete CaCO3 formation with minor MgCO3/FeCO3 is consistent with data for carbonated Archean seafloor plete/pervasive interaction with water is 0.1 km (Earth- basalts (Nakamura & Kato 2004; Shibuya et al. 2012, like) to 50 km. The lower values are more likely, due 2013). Even when not forming carbonates, cations re- to carbonate saturation. The larger value corresponds leased from rocks can control ocean chemistry. Dissolved to a unusually low I/E = 1:1, and a crustal thickness cations (e.g. Na+ and Ca2+) substitute for H+ in the – 100 km – that is large relative to models and rela- ocean charge balance. Removal of H+ raises pH, which tive to Solar System data (Taylor & McLennan 2009; in turn raises the proportion of dissolved inorganic car- O’Rourke & Korenaga 2012; Plesa et al. 2014; Tosi et al. 2017). This might be thought of as including weath- bon that is stored as HCO and CO2− rather than as dis- 3 3 ering of impactors and meteoritic dust. Impact ejecta solved CO2 (Fig. 5). As dissolved CO2 decreases, pCO2 (Sleep & Zahnle 2001) will not contribute cations, be- decreases in proportion (by Henry’s Law). We track only + 2+ cause the ocean will shield the seafloor from impacts and two cations, Na and Ca , because other cations con- the only ejecta will be water. The corresponding cation tribute little to the charge balance of outlet fluid in our abundances (for a 100 km-deep ocean, and neglecting CHIM-XPT runs (Appendix D). Ca-mineral formation) are 0.003 – 1.4 mol/kg Na, and Geophysical arguments about the mass of rock avail- 0.008 – 4 mol/kg Ca. Lumping Na and Ca as moles of able for fluid alteration may be divided into: (i) eruption equivalent charge (Eq), we obtain 0.02 – 9 Eq/kg. history arguments, and (ii) pervasiveness arguments. We assume that rock alteration occurs at T < 650 K (for rea- 3.2. Calculation of Ts(t) and pCO2(t) sons given in Appendix C). (i) Eruption history: Alteration by water is much eas- pCO2(t) is set by sea-surface equilibration between the ier for extrusive rocks (lavas) than for intrusive rocks well-mixed ocean and the atmosphere (Fig. 6). (sills and dikes), because altering fluids are denied entry For each of a wide range of prescribed seawater cation 3 contents and dissolved inorganic C contents, we used by the &10 -fold lower permeability of intrusives (Fisher 1998; Alt 1995). Intrusive:extrusive ratios on Earth are CHIM-XPT to find pH(Tsurf ) and pCO2(Tsurf ). For scattered around an average of ∼5:1 (6.5:1 for oceanic these atmosphere-ocean equilibration runs, we used a to- tal pressure of 20 bars. We found (batch) equilibria in crust) (White et al. 2006; White & Klein 2014). Ex- + 2+ + trusives less than 0.5 km below seafloor – the high- the system H2O-HCO3 –Ca -Na , including formation permeability zone – are the site of most of the alteration of calcite (CaCO3) and portlandite (Ca(OH)2). within oceanic crust on Earth. Altered crust on an ex- We consider the full range of geophysically plausible oplanet could be much thicker if the extrusive pile were cation contents. In §4, we initially describe results for thicker, as is likely for stagnant lid mode. In stagnant lid 3 cation cases: (1) [Na] ≈ 0.5 mol/kg, [Ca] ≈ 0 – higher- mode, each lava erupts at the surface, cools and is rapidly pH, but lacks minerals; (2) [Ca] ≈ [Na] = 0 – a lower-pH altered, and then is buried by later lavas. Extrusives that case; (3) [Ca] = 0.25 mol/kg, [Na] ∼ 0 – higher-pH. Many are completely leached of an element will give an water- of the runs in case (3) form calcite (or, for pH > 9, port- world ocean with an aqueous molality (denoted by square landite). We do not track the fate of Ca-mineral grains, brackets) of which store most of the Ca in our higher-pH equilibrium calculations. Depending on ocean depth and grain size, minerals might stay suspended in the water column, sink 23 z ρ  I −1 [Na] = Na0 cr cr (4) and redissolve, or pile up on the seafloor. Whatever the 1000 zocρw E fate of individual grains, equilibrium with calcite buffers shallow-ocean (and thus atmosphere) CO2 content for where zoc is ocean thickness, zcr is crustal thickness, ocean chemistries that form calcite. 8 Kite & Ford

To find CO2 solubilities, we used previously-published models of CO2 solubility in pure H2O, fit to experi- mental data. Specifically, we interpolated in Table 3 of Carroll et al. (1991) and Table 3 of Duan & Sun (2003), also using the constraint that CO2 solubility = 0 when p = pH2O. For pH2O, we used the formulation in Appendix B of Duan & Sun (2003). We extrapo- lated into the low-T high-P region. CO2 solubility is reduced by salts. For a 1 mol/kg solution at 10 bar and 303K, salting-out reduces solubility by 18% (Duan & Sun 2003). However, we do not include the direct ef- + 2+ fect of Na or Ca on CO2 solubility in our model, reasoning that the pH effect is more important and that the range of C considered in our model is so large that the effect of the ions on CO2 solubility is less im- portant. Using the CO2 solubilities and the carbon- ate system equilibria from the CHIM-XPT calculations, we compute pCO2(Tsurf , [Na], [Ca], [C]). Equipped with pCO2(Tsurf , [Na], [Ca], [C]), we next calculate (for a given planet surface gravity, and fixing atmospheric molecular mass = 44 g/mole) pCO2(Tsurf , [Na], [Ca], cC ). Here, cC is the atmosphere+ocean column C abundance and can (in principle) greatly exceed the ocean C-storage capa- bility. To ameliorate interpolation artifacts, we smooth out each pCO2(Tsurf ) curve using the MATLAB smooth Figure 7. Chart for finding equilibria between ocean chemistry function with a 30 ◦C full-width bandpass. and atmospheric pCO2 (fW = 0.01, a = 1.1 AU). Each green line corresponds to pCO2(Tsurf ) for fixed L∗ (interpolated from Next, we interpolate the Tsurf (L∗, pCO2) results of a Wordsworth & Pierrehumbert 2013; small wiggles are interpola- 1D radiative-convective climate model (Wordsworth & tion artifacts). L∗ spaced at 1 Ga intervals. The other lines cor- Pierrehumbert 2013). Their study neglects clouds and respond to pCO2 as a function of Tsurf , for different assumptions adjusts the surface albedo to a value (0.23) that repro- about ocean cation content: the solid blue lines correspond to a dilute ocean, the dashed red lines correspond to a [Na] = 0.25 duces present-day Earth temperatures with present-day mol/kg ocean, and the dash-dot magenta lines correspond to a CO2 levels. (Comparison to the results of models that [Ca] = 0.125 mol/kg ocean (see §3.3). Line thickness increases include clouds indicate that this assumption will not af- with equal amounts of total atmosphere+ocean C (including C in fect the qualitative trends presented in the current paper; Ca-minerals, if any). Lines of the same thickness have the same to- tal atmosphere+ocean C: {3.5 ×104, 1.1 ×105, 3.5×105, 1.1×106, see §6.5.) We use a log-linear extrapolation to extend 6 −2 −6 3.5×10 } kg m CO2-equivalent C are considered. the fit to pCO2 = 3 × 10 bar. Where Wordsworth & Pierrehumbert (2013) find multiple stable equilibria ment. Lower albedo allows stellar luminosity increase in their 1D model, we use the warmer of their two so- to melt ice-covered surfaces (§6.5). Melting always leads lutions, on the assumption that impacts intermittently to a habitable state, because CO2 does not build up in allow the atmosphere+(wave-mixed ocean) (which have the waterworld approximation and is low upon deglacia- low thermal inertia) to jump onto the warm branch. tion (Yang et al. 2017; Turbet et al. 2017a). If ice cover Once the atmosphere+(wave-mixed ocean) have reached develops later in planetary history, then initially high the warm branch, the deep ocean will gradually adjust to albedos can stay high, but this is very uncommon in our the new, higher temperature. To find ocean-atmosphere model. We evaluate if high-pressure ice phases (ice VI, equilibria, we find {Tsurf , pCO2} combinations that sat- ice VII) are stable at any depth between the ocean sur- isfy L∗(t), [Na], and [Ca] (Fig. 7), as explained in §4.1. face and the seafloor along a pure-liquid-H2O adiabat At each timestep, we check (at all depths within the (Appendix B, Fig. 1.1). We do not include the exso- ocean) for the presence of ice. Low-pressure surface ice lution of CO2, nor partitioning into clathrates, because (ice I) will have steady-state thickness <20 km (due to we conservatively assume HP-ice development means the geothermal heat). We assume albedo of an initially ice- world is no longer habitable. covered surface declines due to low-albedo exogenic con- taminants (meteoritic dust) on a timescale  10 Gyr, 4. RESULTS OF WATERWORLD EVOLUTION MODEL back to 0.23. The value of 0.23 is chosen for conve- All of the results presented in this section are for a nience because it is the surface+clouds albedo used by Sun-like star with (log luminosity)-vs.-time fit to a stan- (Wordsworth & Pierrehumbert 2013). This lowering of dard solar model (Bahcall et al. 2001). For this model, initially high albedo is reasonable because, in our model, L increases by ∼8% Gyr−1. We find that: planets develop surface ice not at all, or very early. Early- ∗ developed ice cover will sweep up debris from planet for- • Waterworld climate is usually stable to T perturbations, mation, which has low labedo (L¨ohne et al. 2008). The although exsolution-driven climate instabilities are pos- low albedo of debris is partly responsible for the Solar sible (§4.1); System observation that all icy worlds have low-albedo • Habitable surface water can persist for many Gyr (§4.2); surfaces except terrains that have been tectonically resur- faced and/or created after the Late Heavy Bombard- • Using geologically and cosmochemically plausible priors in a Monte Carlo calculation, we estimate that ∼1/4 of Habitability of exoplanet waterworlds 9

Figure 8. Time evolution of ocean-surface temperature (Tsurf ) Figure 9. Time evolution of p for a = 1.1 AU, f = 0.01, for a = 1.1 AU, f = 0.01, M = 1 M . Dashed red lines cor- CO2 W W pl ⊕ Mpl = 1 M⊕. Dashed red lines correspond to oceans with respond to oceans with 0.5 mol/kg Na, dash-dotted magenta lines 0.5 mol/kg Na, dash-dotted magenta lines correspond to oceans correspond to oceans with 0.25 mol/kg Ca, and solid blue lines with 0.25 mol/kg Ca, and solid blue lines correspond to dilute correspond to dilute oceans. Grayed-out portions of the lines cor- oceans. Grayed-out portions of the line correspond to parts of respond to parts of the time evolution for which high-pressure ice the time evolution for which HP-ice occurs within the ocean. For occurs at depth within the ocean. For these grayed-out portions, these grayed-out portions, T will be an underestimate to the T will be an underestimate to the extent that HP-ice excludes surf surf extent that HP-ice excludes CO2 from the matrix. The thickness CO from the matrix. The thickness of the line corresponds to 2 of the lines correspond to cC , with thicker lines marking higher c , with thicker lines marking higher c : {3.5 ×104, 1.1 ×105, 4 5 5 6 6 −2 C C cC : {3.5 ×10 , 1.1 ×10 , 3.5×10 , 1.1×10 , 3.5×10 } kg m 5 6 6 −2 3.5×10 , 1.1×10 , 3.5×10 } kg m CO2-equivalent C are consid- CO2-equivalent C are shown for the dilute-ocean case. For the 6 −2 ered. Dotted line marks 398K (highest temperature at which life 0.25 mol/kg Ca and 0.5 mol/kg Na cases, cC = 1.1×10 kg m has been observed to proliferate). Black bar marks 450K (end of CO2-equivalent C plots on the left half of the figure and the smaller habitability). Lines end if pCO2 > 40 bars. values of cC plot on top of one another as the near-vertical line around 7 Ga (only one representative low-cC track for each case is waterworlds have habitable surface water for >1 Gyr shown). Black bar marks 40 bars CO2 (which we deem sufficient (§4.3). for the end of habitability). Lines end if T > 450K.

usually exsolved with increasing T . p increases 4.1. Evolution of individual waterworlds surf CO2 more steeply with Tsurf for the cation-rich cases.

Charting planet evolution. With the waterworld approx- Stability and instability. To see that Tsurf – pCO2 equi- imation, the cations available to the ocean are constant libria are usually stable, consider the point around 50◦C, with time after 0.1 Gyr. Thereafter, Tsurf and the total 1 bar. Suppose we add some energy to the system, per- amount of C in the atmosphere+ocean (including C in turbing Tsurf . This drives CO2 from the ocean to the Ca-minerals, if any) are the sole controls on pCO2. pCO2 atmosphere – moving along the blue line – which will increases with Tsurf due to the low solubility of CO2 in cause more warming leading to a positive feedback on warm water (and related T -dependent shifts in the car- the initial perturbation. However, the value of pCO2 that bonate system equilibria) (Fig. 7). pCO2 = pCO2(Tsurf ) is in equilibrium with the increased surface temperature is a single-valued function for any cation and total-C con- is even higher than the increased CO2, so the climate is centration (Fig. 7). Tsurf = Tsurf (pCO2) greenhouse stable. curves (green lines) can be computed using a climate For P > 1 bar, CO2 exsolution can be a negative (stabi- model given semimajor axis a and L∗(t). To find the lizing) feedback on climate evolution. For example, con- ◦ pCO2 and Tsurf for a given time, we locate the green line sider the point at 40 C, 10 bars in Fig. 7. An increase in corresponding to that star age, then find its lowest-Tsurf L∗ corresponding to 1 Gyr of stellar evolution at constant stable intersection with the appropriate pCO2(Tsurf ) line pCO2 would cause a Tsurf increase of ∼40K. However, the (Fig. 7). climate is constrained to evolve along a curve of constant As L∗ increases with time, Tsurf = Tsurf (pCO2, L∗) in- seawater cation abundance. Therefore, CO2 is exsolved creases for a given pCO2. This is shown by a rightwards from the ocean. Cooling from Rayleigh-scattering per drift of the green lines (greenhouse curves) in Fig. 7. The unit of added CO2(g) exceeds greenhouse warming per green lines are closely spaced at low pCO2, corresponding unit of added CO2(g), so adding CO2(g) cools the planet. to a small Tsurf rise for a given increase in L∗ (Fig. 1). The Tsurf increase is only <10K. This negative feedback This tends to extend the duration of habitable surface is not a geochemical cycle. water, τhab. By contrast the green lines are widely spaced CO2’s solubility in water (for fixed pCO2) increases for ◦ at low pCO2, so a small increase in L∗ corresponds to a Tsurf > (110-160) C. This feedback is mildly stabilizing, rapid Tsurf rise (Fig. 1). The Tsurf increase is usually but rarely important. more than would occur without ocean chemistry (Kitz- For a {pCO2, Tsurf } point to be unstable to mann et al. 2015), because the greenhouse gas CO2 is small Tsurf perturbations, three conditions must be sat- 10 Kite & Ford isfied. (1) ∂pCO2,OC/∂T > ∂pCO2,GH/∂T , where the sub- script OC refers to ocean chemistry curves, and the sub- script GH refers to greenhouse curves. Otherwise, the positive feedback has finite gain, and there is no run- away. (2) ∂pCO2,GH/∂T > 0, and (3) ∂pCO2,OC/∂T > 0 (if (2) and (3) are not satisfied, there is a stabilizing, negative feedback). Exsolution-driven climate instabilities (runaway feed- backs) are possible in the model. For example, consider 6 −2 the Na = 0.25 mol/kg, cC = 3.5 × 10 kg m track (thick red dashed line) from Figs. 8 - 9. At 2.0 Gyr and 1.1 AU, two stable states are possible for this com- ◦ ◦ bination: 37 C and 75 C (Fig. 8). As L∗ increases, both stable states get warmer until at 2.3 Gyr the GH curve no longer intersects the thick red dashed line for ◦ Tsurf < 60 C. The cool branch has disappeared. Be- cause the cool branch has disappeared, the climate sud- denly warms. (Emergence of cold solutions as solar lu- minosity increases is much less common in our model). Figure 10. As Fig. 8, but for fW = 0.1. Evolution of ocean- Increases in pCO2 of up to a factor-of-10 accompanied by surface temperature T , for a = 1.1 AU, f = 0.1, M = 1 M⊕. ◦ surf W pl increases in temperature of up to 100 C can occur in Dashed red lines correspond to oceans with 0.5 mol/kg Na, dash- the model within a single (0.1 Gyr) timestep. dotted magenta lines correspond to oceans with 0.25 mol/kg Ca, How fast would warming occur? The pace of change and solid blue lines correspond to dilute oceans. Note greater per- sistence of HP-ice (gray shrouding) at depth within the oceans. is set by deep-ocean thermal inertia and by the ocean The thickness of the line corresponds to cC , with thicker lines mixing time. If the ocean mixes slowly (i.e. is strat- marking higher cC . Tracks are generally rather similar to in the ified), then the timescale for the instability could be fW = 0.01 case, except that the same climate evolution requires 7 ∼ 10× more cC due to dilution. Dotted line marks 398K (highest as long as the conductive timescale, >10 yr. If in- temperature at which life has been observed to proliferate). Black stead the ocean mixing time is fast, then as each par- bar marks 450K (end of habitability). Lines end if pCO2 > 40 bars. cel of seawater is mixed into the near-surface layer with temperature Tsurf , it will exsolve CO2. Therefore, mate for O(Gyr). The time-integrated change in surface both pCO2 and Tsurf will rise on the fast ocean mixing energy balance (units J) as the consequence of the im- timescale. Rapid temperature change in the wave-mixed pact can easily be 1000× greater than the impact energy. near-surface layer could stress phototrophs. However, for 2 This connection is a new mechanism for a metastable a O(10) W/m change in the surface energy balance, a impact-driven climate. Metastable impact-triggered cli- 100 km-deep ocean will take >100 Kyr to warm. Can mates have previously been proposed using other mecha- this rate of change drive extinction? Microbial popula- nisms (Segura et al. 2012, but see also Wordsworth 2016). tions can make major adaptations on 20-yr timescales (Blount et al. 2008). Therefore, we think that the rate Duration of habitable surface water (τhab). In some cases, of change associated with these instabilities is unlikely habitability can persist for many Gyr (Figs. 8–9). As to seriously threaten whole-ocean microbial extinction, L∗ rises, either Tsurf eventually exceeds 450K, or CO2 provided that the climate conditions at the end-points of exceeds ∼40 bars, either of which we deem sufficient to the instability are habitable. extinguish habitability. (This may be too pessimistic; The existence of instabilities implies that more than §6.4). Different cation cases with the same cC have dif- one surface temperature can satisfy surface energy bal- ferent pCO2(t) trajectories (different colors of the same ance for a given L∗ and cC . Which temperature is cor- thickness in Fig. 9). This is because, for the same total rect? Our planet evolution tracks follow the cool solution amount of C, the fraction of C stored in the ocean differs (as long as it exists) because solar luminosity is initially between cation cases. However, for different cation cases low. Therefore, whole-ocean T will start low and increase the trajectories bunch together at high cC (thick lines of with time due to increasing L∗ – in the absence of other all colors in Fig. 9). This is because for high cC , the ocean perturbations. A candidate perturbation for accessing is saturated and most of the C is in the atmosphere. the warm solutions is an asteroid impact. Very large When (C in atmosphere)  (C in ocean), Catm ≈ Ctotal impacts would be needed, because of the large thermal independent of how much C is in the ocean. The colors inertia of the deep ocean. (A smaller impact can raise the also bunch together for very low cC . For such worlds, the ◦ temperature of the shallow ocean and exsolve some CO2, pCO2 is too low to affect climate until Tsurf > 100 C. but accessing the stable warm branch requires exsolving As a result, such worlds have H2O as the only green- CO2 from the bulk of the ocean). For example, consider house gas. Then, very low cC worlds deglaciate at the one of the largest impact structures on Mars, Hellas. If same time (for a given a), trek quickly across the narrow 27 1/3 of Hellas’ impact energy (total ∼4 × 10 J; Williams “neck” at the bottom of the diagram on Fig. 1, and then & Greeley 1994) goes into heating a 100 km-deep ocean suffer the runaway greenhouse. on a 1 M⊕ planet, ∆Tsurf ≈ 5 K, which is small. Never- In Fig. 8, cases with high cC have longer durations with theless, impacts more energetic than Hellas are possible habitable surface water (τhab). The longest-lived worlds early in planetary history, and if the warm branch is ac- maintain pCO2 in the 0.1-10 bar range for Gyr (Fig. 9). cessed, then the trace of the impact will be seen in the cli- fW = 0.1 worlds (Fig. 10) show greater persistence of HP-ice. This is because fW = 0.1 seafloor pressure Habitability of exoplanet waterworlds 11 is 12 GPa (for Mpl = 1 M⊕), and these high pressures dle panel in Fig. 11). The upper limit is raised if cations favor HP-ice (Fig. 1.1). For a given cC , the evolution is are available (left panel and right panel of Fig. 11), be- colder, because less CO2 is in the atmosphere and more cause they neutralize the carbonic acid. For example, is in the ocean (due to dilution). For high-fw worlds, adding 0.5 mol/kg of positive charges (0.25 mol/kg of because the ocean is the dominant reservoir of C, small Ca2+, or 0.5 mol/kg of Na+) means that an additional fractional changes in the storage of C in the ocean can 0.5 mol/kg of C must be added to get nontrivial CO2 lead to large fractional changes – typically increases – in the atmosphere. For a 100 km deep ocean, this is 6 −2 6 −2 in pCO2. Because ∂pCO2,OC/∂T is larger, the condition 200 bars = 2×10 kg m . 2 × 10 kg m corresponds ∂pCO2,OC/∂T > ∂pCO2,GH/∂T occurs for a wider range to the vertical offset between the habitability optimum of {pCO2,Tsurf } parameter space and thus exsolution- in the middle panel, versus the habitability optimum driven climate instabilities can have larger amplitude. in the two side panels, of Fig. 11. 2 × 106 kg m−2 The initial pH of the habitable surface water ranges also corresponds to the vertical offset between the up- from 3 to >11. In our model, as atmosphere pCO2 rises per limit for habitability in the middle panel, versus the during the interval of habitable surface water, ocean upper limit for habitability in the two side panels, of pH falls (Zeebe 2012), typically by < 1 pH unit. For Fig. 11. Neutralization by cations also explains why τhab {pCO2, Tsurf } combinations that allow habitable surface is brief (and uniform) for low C contents when cation water, pH is higher for the ‘0.25 mol/kg Ca’ case and content is high. For these worlds, pCO2 is so low (be- highest of all for the ‘0.5 mol/kg Na’ case. The relative cause pH is so high) that these planets effectively have effectiveness of Na relative to its charge-equivalent in Ca H2O + N2 atmospheres. Examples of evolutionary tracks is due to the partitioning of Ca into Ca-minerals. for such worlds include the nearly-vertical tracks to the In summary, ocean chemistry is key to waterworld cli- right of Fig. 8. Pure-H2O atmospheres lead to planets mate evolution. This is because (1) ocean pH sets the having habitable surface water for < 1 Gyr if they orbit ocean-atmosphere partitioning of C, which in turn sets Sunlike stars, according to the 1D model of Goldblatt the offsets between the y-axis positions of lines of the (2015). same color (cation abundance) but different thickness τhab is longest for semimajor axes a ∼1.1 AU – defining (cC ) in Figs. 7-10; (2) carbonate-system equilibria are “bull’s-eyes” in Fig. 11. This is because habitable surface T -dependent, which causes the upwards slope of the col- water is available from t = 0 at the inner edge of the HZ, ored lines in Fig. 7-10. but the runaway greenhouse comes soon. For large a, the surface is frozen until near the end of the model run (and 4.2. Controls on the duration of the end of the main sequence, which we decree to occur habitable surface water (τhab) at 10 Gyr for the Sunlike host star, arrives swiftly). For a ∼ 1.5 AU, only a narrow range of high CO2 permits Fig. 11 shows, for fW = 0.01, Mpl = 1 M⊕ water- worlds, the onset time, shutoff time, and duration of hab- habitability, corresponding to the wing-shape extending itable surface water. In each case, nonzero onset times to the top right on all panels. CO2 condensation (Turbet (thin solid lines in Fig. 11) correspond to meltback of an et al. 2017a) might clip this wing of high-a habitability. early-established ice lid on the ocean. Shutoff times (thin Only near 1.1 AU does habitable surface water both be- dash-dot lines in Fig. 11) correspond to either the run- gin early and end late, corresponding to the “bull’s-eyes.” away greenhouse or to excessive (>40 bar) p . With On a 10×-deeper ocean (fW = 0.1; Fig. 12), the hab- CO2 itability optimum has moved to 10× larger values of C. increasing cC , for all panels in Fig. 11, τhab (thick solid lines in Fig. 11) rises and then falls over an interval of This is because more C is needed to overwhelm the ocean 6 −2 2 sinks. The range from “minor CO2 to 40 bars” is still ∼ 10 kg m , i.e. O(10 ) bars of CO2-equivalent. This only 4 × 105 kg m−2 C for fixed cations, so the range of C exceeds the O(10) bar pCO2 range for optimum habit- that gives long durations of surface liquid water appears ability in Fig. 1. That is because around the habitability narrower on a log scale. The upper limit for habitability optimum, as C is added to the system, most of the ad- has also moved higher, but only by a factor of 2. This is ditional C is partitioned into the ocean and does not because [C]-rich tracks that reach 40 bars at T < 100 ◦C contribute to pCO2 (Archer et al. 2009). always have HP ice for fW = 0.1, and so do not count The location of the upper limits on habitability can as habitable (thick gray lines in bottom left of Fig. 10). be understood by reference to a basic “bucket model” of The strong a preference for fW = 0.01 is more subdued C partitioning (Fig. 13). Starting with a small cC and for fW = 0.1. This is because almost all fW = 0.1 worlds adding C, pCO2 is initially low until C overwhelms rock- start with HP-ice due to the deeper ocean (Fig. 1.1), and sourced cations and the ocean becomes acidic. Once the such worlds are not counted as habitable in Fig. 12 until ocean is acidic, additional C spills into the atmosphere temperature has risen sufficiently to melt all the HP-ice. and pCO2 inevitably becomes high. Supposing the ocean storage capacity for C to be fixed at ∼0.5 mol/kg C (in 4.3. Summary and Monte Carlo results reality capacity will rise with pCO2; Eqn. 1), and for So far we have been considering three specific cation- pH low enough that all aqueous C is stored as CO2(aq), abundance cases. Now we consider a much wider range. then ocean CO2-equivalent content is 20 g/kg – for a Fig. 14 shows how τhab depends on cC and the extent 6 −2 100 km deep ocean, 3 × 10 kg m . If extra C goes into of rock-water interaction for a = 1.1 AU, fW = 0.01. For 5 −2 the atmosphere, then an additional 4 × 10 kg m leads high cC (black circles in Fig. 14), C swamps all available to an uninhabitably hot surface. Although the basic oceanic and crustal sinks and so C piles up in the atmo- “bucket model” is an (over-)simplification of carbonate sphere (Fig. 13), leading to sterilizing surface tempera- system chemistry, this explains the ∼106 kg m−2 upper tures When cations released from rock dominate ocean limit for habitability for the Na-free, Ca-free case (mid- chemistry, pH is high, and almost all C is sequestered as 12 Kite & Ford

Figure 11. Waterworld τhab (duration with habitable surface water) diagrams for fw = 0.01, Mpl = 1M⊕. Column-masses of C are in CO2 equivalent. Thick contours show the duration with habitable surface water, in Gyr. They are spaced at intervals of 1 Gyr, except for the outer dotted contour, which corresponds to a duration of 0.2 Gyr. Thin solid contours correspond to the time of onset of habitable surface water (i.e., time after which ocean has neither HP-ice nor surface ice), spaced at intervals of 1 Gyr. Worlds to the right of the leftmost thin solid contour in each panel are “revival worlds” (worlds that start with surface ice, and subsequently deglaciate). Thin dashed contours correspond to the time of cessation of habitable surface water (i.e., either pCO2 > 40 bars, or Tsurf > 450K), spaced at intervals of 1 Gyr. The small wiggles in the contours, and the indentations at ∼3×107 kg m−2, are interpolation artifacts.

Figure 12. As Fig. 11, but for fw = 0.1.

that permits habitable surface water is widest for pCO2 ∼ 1 bar (Fig. 1). Durations dwindle for lower [C] and van- ish entirely for higher [C] (excessive pCO2 and/or Tsurf ). When cations in the ocean balance [C] (black horizontal line), the maximum in habitable-surface-water duration occurs at larger C. We suppose that the altered-rock layer has the composition of mid-ocean ridge basalt, that it is completely leached of the charge-equivalent of its Na and Ca content, and that the leached thickness is log- uniformly distributed from 0.1 km to 50 km (§3.2). For the purposes of Figs. 14-15 we assume that carbonates don’t form; if they did form, then a factor of ∼2 more C would be needed to neutralize the cations. Equilibrium with minerals (e.g. CaCO3, NaAlSi2O6) might buffer ocean cation content to <0.5 mol/kg, which roughly cor- responds to the middle red line in Fig. 14. Most points Figure 13. A basic “bucket” model of p . above the middle red line in Fig. 14 have short τhab. CO2 Therefore, our decision to allow cation contents >0.5 2− − CO3 ion, HCO3 ion, or CaCO3 (Kempe & Degens 1985; mol/kg is geochemically less realistic, but conservative Bl¨attleret al. 2017). Therefore little C remains for the for the purpose of estimating τhab on waterworlds. atmosphere, and so pCO2 is minimal and has little role in A Monte Carlo exercise illuminates habitable- setting planet climate, and the planet has habitable sur- surface-water duration for a hypothetical ensemble face waters for only <1 Gyr (dark blue disks at the top of Mpl = M⊕, fW = 0.01, a = 1.1 AU worlds of Fig. 14). For cation-poor oceans (lower part of plot), (Fig. 15). Suppose that log(cC ) is uniformly distributed τhab is longest for [C] that gives pCO2 ∼ 1 bar at habit- (independent of the leached thickness) from -5 to -3 able temperatures, as expected because the range of L∗ Habitability of exoplanet waterworlds 13

Figure 14. To show controls on τhab, the duration of habit- able surface water. Color scale corresponds to τhab (Gyr) for a = 1.1 AU, fW = 0.01, M = 1 M⊕. Black circles are never- habitable worlds. Red lines show the cation content (mEq/kg) for complete leaching of both Na and Ca – with no mineral formation – from a basalt column 50 km thick (top line), 5 km thick (mid- dle line), or 0.1 km thick (bottom line). The top and bottom red lines are the bounds for cation content used in the Monte Carlo calculations. The dotted grey line is the Na from condensation of a 100-bar steam ocean, assuming a volume mixing ratio in that steam ocean of 0.002 NaCl(g) (Lupu et al. 2014). The vertical dashed lines are the bounds for C abundance used in the Monte Carlo calculations. (0.3% - 30% of water content; C/H> 1 is cosmochem- ically implausible)8 (§3.2). For these cosmochemi- cally and geophysically reasonable priors, the dura- tion of habitable surface water is shown in Fig. 15. Fig. 15’s lower panel shows that C has a negative or neu- tral effect on τhab for slightly more than half of the Monte 1 Carlo worlds. About /4 of worlds never have habitable Figure 15. Results of Monte Carlo calculations. Upper panel: 1 surface water, due to excessive initial pCO2. Another /4 Color scale corresponds to the duration of habitable surface water, have pCO2 so low, due to efficient rock leaching, high-pH τhab (Gyr). See text for justification of priors. Open black circles 2− − are never–habitable worlds. For M = 1 M⊕, a = 0.01, fW = 0.01. oceans, and storage of C as CO3 ion or HCO3 ion, that Middle panel: The same, but for fW = 0.1. Note change in color CO2 has minor climate effect. As a result, the (short) scale. Lower panel: Output sorted by the duration of habitable duration of habitable surface water is about the same surface water. Solid line, fW = 0.01; dash-dot line, fW = 0.1. as for an N2+H2O atmosphere and < 1 Gyr. The re- 1 maining /2 of worlds have longer τhab due to CO2. In challenge their habitability. ∼9% of cases, the waterworlds stay habitable for longer The area under the curves in the lowest panel of Fig. 15 than the age of the Earth. is a measure of the efficiency of cycle-independent plan- Fig. 15 also shows the results for fW = 0.1. Durations etary habitability at sustaining habitable surface water of habitable surface water are shorter, because HP-ice relative to a hypothetical geochemical cycle that main- can only be avoided for a narrow (steamy) range of tem- tains pCO2 at the perfect value for habitability. This is peratures. Reducing the maximum temperature for life because, even if habitability is maintained by geochemi- from 450 K to 400 K would reduce τhab to < 1 Gyr on cal cycles, maximum possible τhab for a = 1.1 AU, Mpl fW = 0.1 worlds. For deeper oceans, fewer worlds are = 1 M⊕ is 7 Gyr. After 7 Gyr, the runaway greenhouse sterilized by too-high CO2 values: for these worlds, the will occur even for pCO2 = 0 (Caldeira & Kasting 1992; solution to pollution is dilution. Wolf & Toon 2015). Our results are sensitive to the mean value of cC /fW . The effect of C on habitability in our model is positive −3 (Fig. 15, lowest panel). First, our Monte Carlo calcu- If cC often exceeds 10 , then a greater fraction of wa- terworlds will have thick CO -rich atmospheres that will lations suggest that the mean τhab is comparable in the 2 random-allocation-of-C case relative to a C-starved case. 8 This prior likely overweights small values of cC , because for Second, the maximum τhab is increased (Carter 1983). small values of cC , C is likely to be stored mainly in the silicate These gains outweigh the losses due to planets that form mantle (Hirschmann 2016). with too much C for habitable surface water. In addi- 14 Kite & Ford tion, life requires C, so very low [C] in the ocean might within a system rapidly damp to a level where damping challenge the origin of life and could require organisms to is balanced by N-body excitation from neighbors. Based develop C-concentrating mechanisms in order to survive on previous experience, this initialization helps to avoid (Kah & Riding 2007). a phase of extremely rapid collisions before the system is able to relax to a physically plausible set of inclinations 5. PLANET ASSEMBLY MODEL and eccentricities. The three remaining angles (argu- The purpose of our N-body simulations (§5.1) is ment of pericenter, longitude of ascending node, mean to provide a physically well-motivated ensemble of anomaly) were drawn from uniform distributions (0 ↔ {a, Mpl, fW } tuples as input for the waterworld evo- 2π). lution model. We choose initial conditions that lead to While the gas-disk is present, orbital migration was large-scale migration, as this is the easiest way to form parameterized following Eqns. 19 & 20 from Zhou & Lin habitable-zone waterworlds. In §5.2, we will apply a (2007). The gravitational acceleration due to tides from volatile tracking code to the mass and orbit histories out- the gas disk are proportional to: (a) the difference in the put by the N-body simulations. Embryos are assigned an velocity of the planet (or embryo) from the circular veloc- initial volatile mass percentage that ramps from zero well ity at the planet’s current position (i.e., proxy for the gas inside the snowline to fW,max well beyond the snowline. velocity), (b) the planet-star mass ratio, (c) a 1/a2 term The embryos evolve via migration and mutual gravita- that accounts for a a−3/2 power-law dependence of the tional scattering, and develop into planets, some of which gas-disk surface density, and (d) a time-dependent factor sit within the HZ. Pebbles (Sato et al. 2016; Levison et that exhibits first slow dispersal, and then rapid disper- al. 2015) and planetesimals are not tracked. The volatile sal due to photoevaporation. This prescription results in content of the forming planets is tracked as the embryos orbital migration, eccentricity damping, and inclination merge. damping that have self-consistent timescales (Zhou & Lin 5.1. N-body model ensemble 2007). For each of our simulations, the gas-disk mass ini- tially undergoes exponential decay with a characteristic We perform N-body simulations including simple grav- timescale of 2 Myr. Starting at 3 Myr, the gas-disk decay itational interactions with a smooth, 1-D gas disk with −3/2 accelerates due to photoevaporation, so the exponential a surface density proportional to r , but not includ- decay rate is decreased to 50,000 years. After 4 Myr, the ing the back reaction of disk pertubations on the planets. gas disk mass has essentially dispersed, so the gas-disk This results in a simple strong-migration scenario, similar mass is set to zero and the simulation becomes a pure to that in many previous studies (e.g. Cossou et al. 2014; N-body integration. Ogihara et al. 2015; Sun et al. 2017; Izidoro et al. 2017). Following gas disk dispersal, planets continue to scat- For the purposes of this study, the N-body simulations ter and collide with one another. We consider planets as serve to provide illustrative migration and collisional his- having collided when they come within a radius corre- tories. Therefore, details of the N-body simulations are sponding to a density of 0.125 × that of Mars. Collisions not essential; some readers may choose to skip to §5.2. are assumed to be perfect mergers in the N-body code. We employ the REBOUND N-body code (Rein & Liu Of our simulations, 29% of the runs resulted in one 2012) and its IAS15 integrator (Rein & Spiegel 2015), or more planets in the HZ, based on the zero-age main with additional user-defined forces to represent gravita- sequence luminosity of the Sun (∼0.7× present solar lu- tional tidal drag from the disk (Zhou & Lin 2007). Each minosity). We set aside N-body simulations where no simulation is initialized with dozens of planerary embryos planets survived exterior to the HZ, as the results of those near or beyond the H2O snowline, orbiting a solar-mass simulations may be impacted by edge effects (i.e., if em- star. No gas giants exist at the start of our simula- bryos formed beyond the outer end of our initial set of tion, consistent with the wetter-than-solar-system cases embryos). In some cases, multiple planets in the HZ at we model (Batygin & Laughlin 2015; Morbidelli et al. the end of the simulation have semimajor axes differing 2016). The following parameters are varied across simu- by less than <0.05 AU from one another. In these cases, lations: the total mass in embryos (from 10 to 40 M⊕); we merge such planets, as they are expected to collide if the outer edge of the disk (from 3 to 10 AU); and a scale we were able to extend each of the n-body simulations. factor for the mass of the gas disk relative to a fidicuial We do not include the loss of water due to these (final) minimum mass solar nebula (from 2-to-16). giant impacts. After applying these filters, we are left For each set of model parameters, we generate with ∼30 habitable-zone planets. multiple sets of initial conditions, resulting in a to- tal of >145 N-body runs. Each embryo’s initial mass is 5.2. Volatile-tracking model drawn uniformly between 0.5 and 2 Mars masses, in se- Embryo H2O content is set as follows. We represent quence until the total embryo mass exceeds the targeted the cumulative effects of the evolving H O-ice snowline total embryo mass. Then, all of the embryo masses for 2 on the water mass fraction of planetary embryos fW,o a given planetary system are rescaled, so that the total (Hartmann et al. 2017; Piso et al. 2015) by embryo mass equals the target total embryo mass. The initial semi-major axes were drawn from a power-law dis- tribution with a power-law index of -1.5. Orbits are ini- fW,o = fW,max (a − asl/asl) , asl < a < 2asl tialized with modest eccentricities and low inclinations fW,o = fW,max, a > 2asl (5) (relative to the plane of the gas disk). The initial orbital f = 0, a < a eccentricities and inclinations are drawn from Rayleigh W,o sl distributions with a Rayleigh parameter of 0.1 or 0.1 where asl = 1.6 AU is an anchor for the evolving snow- radians. In practice, the eccentricities and inclinations line (Mulders et al. 2015). For each N-body simulation, Habitability of exoplanet waterworlds 15

Figure 16. Example output from our model showing stages in planet growth. Color scale shows H2O fraction (fW ). Red rings mark planets that initially have habitable surface water. Green rings mark habitable-zone planets with initially frozen surfaces. Disk size shows planet mass. Planets migrate inwards for the first few Myr of the simulation subject to parameterized, imposed mi- Figure 17. Model output for maximum embryo water con- gration torques. tent (fW,max) = 20%, showing the HZ planets generated by an ensemble of N-body simulations. Disk size increases with Mpl, we consider fW,max = {50%, 20%, 5%, 1%}. The highest which ranges from 0.1 to 1.5 M⊕. For each planet, 1000 draws are value, 50%, corresponds to theoretical expectations from made from a cosmochemically-reasonable range of cC and (inde- pendently) from a geophysically-reasonable range of cation abun- condensation of solar gas (Ciesla et al. 2015). Interme- dance (§3.2), and the resulting durations of habitable surface water diate values match density data for Ceres (Thomas et are computed using the methods of §4. The disk colors summarize al. 2005) and inferences based on Earth’s oxidation state the outcomes. The yellow planet at a = 0.85 AU has seafloor pres- (Rubie et al. 2015). 5% corresponds to meteorite (car- sures too low to suppress volcanism. Dark blue colors correspond to τhab ≥ 1 Gyr. Zero worlds have seafloor P > 8 GPa, so no bonaceous chondrite) data. Why are the carbonaceous worlds are colored purple. chondrites relatively dry? One explanation involves heat- ing of planetesimals by short-lived radionuclides (SLRs; Abe 2005), using the relative-velocity vectors from the 26Al and 60Fe), leading to melting of water-ice. Meltwa- N-body code. The dependence on relative-velocity of ter then oxidizes the rocks, and H returns to the nebula fractional ocean loss is taken from Stewart et al. 2014. 2 Because the parameterisation of Stewart et al. 2014 is (e.g., Young 2001; Rosenberg et al. 2001). Because most −4 protoplanetary disks probably had fewer SLRs than our for fW ∼ 10 , and ocean loss efficiency decreases with own (Gaidos et al. 2009; Gounelle 2015; Lichtenberg et ocean thickness (Inamdar & Schlichting 2016), this is al. 2016; Dwardakas et al. 2017; see also Jura et al. 2013), conservative in terms of forming waterworlds, because- More recent calculations find lower ocean-loss values fW,max likely varies between planetary systems. Another explanation of the dryness of carbonaceous chondrites re- (Burger et al. 2017). Upon embryo-embryo or embryo- lies on Jupiter’s early formation (Morbidelli et al. 2016); planet collision (Golabek et al. 2017; Maindl et al. 2017), gas giants of Jupiter’s size and semimajor axis are not Fe-metal cores are assumed to merge efficiently. We do not track protection of H O from giant impacts by disso- typical. The range of fW,max we consider corresponds 2 to processes occurring during growth from gas+dust to lution within deep magma oceans formed during earlier giant impacts. This is also conservative in terms of form- embryos. We neglect H2O loss by XUV-stripping at the pre-embryo stage (Odert et al. 2017). ing waterworlds. Growing planets shed water during giant impacts. For volatile-rich planets, at the embryo stage and larger, the 5.3. Results most plausible mechanism for losing many wt% of water Planet assembly is gentle in our model. Most planet- is giant impacts. The shift to giant-impact dominance forming collisions occur while disk gas maintains low can be seen by comparing Figs. 14-15 of Schlichting et planet eccentricities and inclinations. As a result, v∞ is al. 2015, and extrapolating to the smallest volatile layer small, and these early collisions have individually-minor 8 −2 considered in this paper (10 kg m ). Such deep volatile effects on planet mass. When fW ≥ 0.15, an individual layers greatly inhibit volatile loss by r<50 km projec- giant impact does not remove a large percentage of the tiles, and blunt the escape-to-space efficiency of larger initial planet volatile endowment (Fig. 16). projectiles. (For volatile-poor planets, by contrast, small fW is negatively correlated with Mpl due to the larger impacts are extremely erosive (Schlichting et al. 2015).) number of devolatilizing giant impacts needed to assem- Combined with the expectation that most-mass-is-in- ble larger planets. For a given Mpl, worlds assembled the-largest projectiles, erosion by giant impact is the via collisions between equal-mass impactors have higher main water loss process. fW than worlds that are built up piecemeal (one em- H2O is attrited by individually-tracked giant impacts bryo added a time). This is because piecemeal assembly (Marcus et al. 2010a; Stewart et al. 2014; Genda & requires a greater number of giant impacts (Inamdar & 16 Kite & Ford

Schlichting 2016). There is no obvious trend of fW with a ble feedbacks, not included in our code, that might sta- (or Mpl with a) within the HZ. bilize climate. The final fW values are sensitive to fW,max, as ex- pected (Mulders et al. 2015; Ciesla et al. 2015). For 1. Carbonate dissolution consumes CO2 (sic; Zeebe a given fW,max, seafloor pressures are bunched because 2012). This is because (for circumneutral pH) one mole both the weight of unit water, and the number of water- 2+ − of Ca charge-balances two moles of HCO3 , whereas removing giant impacts, increase with planet mass. None CaCO3 pairs one mole of Ca with only one mole of C of the fW,max = 0.05 or fW,max = 0.01 planets have (Fig. 5). Dissolution of carbonate by increases in tem- seafloor pressures >1 GPa. By contrast, 97% of the perature (Dolejˇs& Manning 2010) will consume CO2, fW,max = 0.2 worlds have 1-8 GPa seafloor pressure. a climate-stabilizing feedback. This feedback is cur- Finally, of the fW,max = 0.5 worlds, 7% have seafloor tailed on Earth by the build-up of an lag of insoluble pressure <1 GPa, 3% have seafloor pressure ∼ 6 GPa, sediment, for example as continent-derived sediment. and 90% have seafloor pressure > 8 GPa. These rela- Insoluble sediments would be in short supply on wa- tively low seafloor pressures are the consequence of low terworlds, and so the carbonate-dissolution feedback surface gravity: Mpl averages 0.3 M⊕ for this ensemble. could be stronger on waterworlds. To calculate gravity, we follow Valencia et al. (2007) and 0.27 let Rpl/R⊕ = (Mpl/M⊕) . If more massive planets 2. Pervasive aqueous alteration lowers the density of the had emerged in our simulations, then we would expect upper-crust. Low altered-layer density may cause ris- seafloor pressures to be higher for a given fW,max. ing magma to spread out beneath the altered layer, We injected the {a, Mpl, fW } tuples from the and crystallize as an intrusion that has low permeabil- fW,max = 0.2 planet assembly run into the waterworld ity and so is resistant to alteration (instead of extrud- evolution code, varying cC and cation-leaching using the ing as a lava that has high-permeability and is easily same procedure and limits as in §4. The results (Fig. 17) altered). This negative feedback on aqueous alteration show that habitable surface water durations > 1 Gyr oc- might limit cation supply. cur for ∼20% of the Monte Carlo trials. τhab >4.5 Gyr occurs for ∼5% of Monte Carlo trials. Only ∼1/3 of trials 3. Rejection of C by crystallizing HP-ice rejection is a never have habitable surface water. Never-habitable out- negative feedback on cooling. The strength of this ckomes can correspond to always-icy surfaces (especially feedback will depend on the extent to which C is in- corporated into clathrates (Levi et al. 2017). If HP-ice near the outer edge of the HZ), or to pCO2 40 bars. & enshrouds the seafloor before much crust production Never-habitable outcomes become less likely as fW in- creases, mainly because of dilution. These results are has occurred, then pCO2 will be higher than without HP-ice because the HP-ice will inhibit cation delivery illustrative only because mean (and median) Mpl for this to the ocean. This will raise surface temperature and ensemble is 0.3 M⊕. The lower gravity of 0.3 M⊕ worlds accelerate the melting of HP-ice. relative to our 1 M⊕ reference will move the inner edge of the HZ further out than is assumed in Fig. 17 (Kop- parapu et al. 2014). 4. High C concentration lowers pH. This favors rock leaching, which is a negative feedback on pCO2 (Schoo- 6. DISCUSSION nen & Smirnov 2016). But because water:rock ratio is effectively a free parameter over many orders of magni- 6.1. All small-radius habitable zone planets are tude, the effect of W/R is more important in our model potentially waterworlds (Fig. 18). Moreover, high-pH NaOH solutions corrode Waterworlds cannot be distinguished from bare rocks silicate glass, a positive feedback. using radius/density determinations. Even if radius and density measurements were perfect, a small increase in 5. It has been suggested that seafloor basalt carbonitiza- core mass fraction could offset a 100 km increase in ocean tion is an important C sink on Earth, is temperature- depth (Rogers & Seager 2010). Indeed, scatter in core dependent, and so might contribute to Earth’s climate mass fraction is expected for planets assembled by gi- stability (Coogan et al. 2016). ant impacts(Marcus et al. 2010b). The best-determined planet radius is that of Kepler-93b; 1.481±0.019 R⊕ These feedbacks might extend τhab beyond the already- (Ballard et al. 2014); an error of 120 km in radius. long durations calculated in §4. Yet a 120 km deep ocean is enough to be in the cycle- independent planetary regime discussed in this paper, 6.3. Cycle-dependent versus cycle-independent and a 0 km deep ocean is a bare rock. Small uncer- planetary habitability tainties in star radius, or in silicate-mantle composition Given a τhab look-up table, the number of planet-years (e.g. Dorn et al. 2015; Stamenkovi´c& Seager 2016) can of habitable surface water enabled by cycle-independent also affect retrievals. Therefore, all small-radius HZ plan- planetary habitability on waterworlds (Tnocycle,ww) can ets are potentially waterworlds, and this will remain the be estimated: case for many years (Dorn et al. 2017; Unterborn et al. 2017). η⊕N∗ X 6.2. Climate-stabilizing feedbacks? Tnocycle,ww = τhab,i(fW,i, cC,i,Mpl,i, ...) (6) The main climate feedback in our code is the decrease i=1 in CO2 solubility with increasing temperature, which is where N∗ is the number of Sunlike stars in the galaxy, usually destabilizing (§4.1). We now speculate on possi- and η⊕ is the number of HZ small-radius exoplanets per Habitability of exoplanet waterworlds 17

Sunlike star. How does Tnocycle,ww compare to the num- vatively low upper limit. Above 70 bars and 300K, CO2 ber of planet-years of habitable surface water maintained is a supercritical fluid. Supercritical CO2 is an excel- by geochemical cycles (Tcycle)? lent organic solvent and bactericidal agent. However, within the ocean, life might persist. We consider near- Tcycle ∼ η⊕N∗( xIC xevo ) 7 Gyr (7) sea-surface habitability in this paper, but note that life proliferates at 1 GPa (Sharma et al. 2002). Here, xIC is the fraction of small-radius HZ exoplan- Extreme pH values destabilize simple biological ets with suitable initial conditions for habitability main- molecules, but life on Earth solves these problems (COEL tained by geochemical cycles, xevo accounts for planets 2007). The pH requirements for the origin of life might where geochemical cycles initially maintain habitability be more restrictive. but subsequently fail to do so, and 7 Gyr is a typical Life-as-we-know-it requires nutrients other than C, e.g. maximum duration of habitable surface water (before the P. Yet it is unclear to what extent life requires resupply runaway greenhouse) for a planet initialized at random of such nutrients (Moore et al. 2017). To the contrary, a within the HZ of a sunlike star. given an initial nutrient budget and a sustained source of Despite the familiarity of habitability maintained photosynthetic energy, a biosphere might be sustained by by geochemical cycles, it is possible that Tcycle  heterotrophy, i.e. biomass recycling. Biomass recycling Tnocycle,ww. This is because xIC and xevo are very is easier for waterworlds that lack ocean-mantle geochem- uncertain. This uncertainty is caused by the diffi- ical cycling, because biomass cannot be buried by silici- culty of predicting geochemical cycles involving habit- clastic sediment, nor subducted by tectonics. If biomass able environments using basic models. This is trace- recycling is incomplete (Rothman & Forney 2007), then able to the low temperatures of habitable environments, to stave off productivity decline will require nutrient re- which bring kinetic factors – such as grain size, per- supply. Nutrient resupply will be slower (relative to meability, catalysts, and rock exposure mechanisms – Earth) on waterworlds that lack ocean-mantle geochem- to the fore (White & Brantley 1995). It is also diffi- ical cycling, because land is absent (so subaerial weath- cult to forensically reconstruct the processes that buffer ering cannot occur), and volcanism is absent or minor. Earth’s pCO2 (Broecker & Sanyal 1998; Beerling 2008). Moreover, the larger water volume will dilute nutrients. A negative feedback involving geochemical cycles prob- In our model, nutrients are supplied by water-rock in- ably buffers the post-0.4 Ga atmospheric partial pres- teractions at the seafloor early in planet history, with sure of CO2 (pCO2) in Earth’s atmosphere (e.g., Zeebe limited nutrient resupply thereafter (by slow diffusive & Caldeira 2008; Stolper et al. 2016). The effectiveness seafloor leaching, or minor ongoing volcanism). Resup- and underpinnings of these feedbacks on Earth are poorly ply might be shut down entirely by HP-ice. However, understood (Edmond & Huh 2003; Maher & Chamber- if HP-ice both convects and is melted at its base (as on lain 2014; Galy et al. 2015; Krissansen-Totton & Catling Ganymede?), then nutrient supply to the ocean may con- 2017). Therefore, it is hard to say how often these feed- tinue (e.g., Kalousov´aet al. 2018). Alternatively, habit- backs break down or are overwhelmed due to the absence able surface water can be fertilized by bolides and in- of land, supply limitation / sluggish tectonics, inhibi- terplanetary dust, just as Earth’s seafloor is nourished tion of carbonate formation due to low pH or to SO2, by occasional whale-falls. On a planet with habitable contingencies of tectonic or biological evolution, and so surface water, chemosynthesis is greatly exceeded as a on (Abbot et al. 2012; Kopp et al. 2005; Edwards & potential source of energy by photosynthesis. If pho- Ehlmann 2015; Bullock & Moore 2007; Foley 2015; Fo- tosynthesis occurs, then (given the likelihood of nutri- ley & Smye 2017; Halevy & Schrag 2009; Galbraith & ent recycling), it has not clear how potentially observ- Eggleston 2017). Without a mechanistic understanding able parameters would scale with geological nutrient flux. of planet-scale climate-stabilizing feedbacks on Earth, we Therefore, it is hard to make a testable prediction about cannot say whether these feedbacks are common among the nutrient budgets of water-rich exoplanets that have planets, or very uncommon (Lacki 2016). habitable surface water. 6.4. Limits to life 6.5. Uncertainties Is 450K a reasonable upper T limit for habitability? Astrophysics. It would be difficult to gather enough wa- 450K exceeds the highest temperature for which life has ter to form a waterworld if habitable-zone planets form been observed to proliferate (398K, Takai et al. 2008). without a major contribution of material from beyond However, no fundamental barriers have been identified to the snow-line. If equilibration of metal with atmophiles life at >400K (COEL 2007). Theory tentatively suggests during giant impacts is inefficient (Rubie et al. 2015), 423-453K as the upper T limit for life in general (Bains then it is difficult to sequester C in the core. This would et al. 2015). If we lower the limit to life to 400K, then our lead to an atmosphere+ocean that inherits the C/H an- model fW = 0.1 oceans will almost always have HP-ice at ticipated for nebular materials, which is more C-rich than 0 2+ times when their surface temperature is habitable (Fig. considered here. The reaction Fe + H2O → Fe O + 10). Earth’s Tsurf may have been >350K at >3 Ga based H2 could also raise C/H. on isotopic data (e.g. Tartese et al. 2017) and on the T Our results are sensitive to the assumption that tolerance of resurrected ancestral proteins (Akanuma et initially-ice-coated worlds gather enough meteoritic de- al. 2013). bris during the late stages of planet accretion to lower Is 40 bars a reasonable upper pCO2 limit for habitabil- albedo. If albedos of initially-ice-coated worlds all stay ity? Wee use this upper limit mainly because fine-tuning so high that deglaciation cannot occur, the only habit- of L∗ is increasingly required to avoid uninhabitably high able waterworlds will be those that are unfrozen for zero- Tsurf above this value. Therefore, 40 bars is a conser- age main sequence stellar luminosity (a . 1.1 AU). For 18 Kite & Ford a . 1.1 AU, the runaway greenhouse occurs relatively We neglect Cl. We do so because Cl is geochemically early in stellar main sequence evolution, so if deglacia- much less abundant than Na and C (Sleep et al. 2001; tion cannot occur, most waterworlds with habitable sur- Grotzinger & Kasting 1993; Sharp & Draper 2013; Clay face water will orbit stars younger than the Sun. Be- 2017). Because we neglect Cl, our model understates the cause many HZ planets have a & 1.1 AU, a complete acidity of low-cC worlds. Therefore, the vertical stripe of prohibition on deglaciation would lower the number of “optimal cC ” is drawn at too high a value of cC . waterworlds with habitable surface water by a factor of Our modeled water-rock interactions, which use fresh- ∼2. water as the input fluid, raise pH due to release of (for + 2+ Given our assumption that our worlds lack H2, it is example) Na and Ca (Appendix D) (MacLeod et al. self-consistent to assume waterworlds have CO2 (and 1994). Na is released from seafloor basalt (Staudigel & not CH4) atmospheres. If H2 nevertheless remains, then Hart 1983) in our model, consistent with experiments it could fuel early hydrodynamic escape of H2O, and aid (Gysi & Stef´ansson2012) and Na-leaching of preserved Fischer-Tropsch type synthesis of organics (Gaidos et al. 3.4 Ga seafloor crust (Nakamura & Kato 2004). More- 2005). We assume a fixed pN2, ∼0.8 bar, but variation over, on waterworld seafloors, jadeite is the stable Na- in the abundance of N volatiles may be important, espe- hosting phase, and (at least for T > 400C) the albite- cially if planets draw material from a & 10 AU (Maroun- jadeite transition is a maximum in [Na] (Galvez et al. ina et al. 2017b; Atreya et al. 2010). NH3 is a freezing- 2016). Na release into waterworld oceans is consistent point depressant, and N2 affects Tsurf by pressure broad- with Na uptake at Earth hydrothermal vents (Rosen- ening, Rayleigh scattering, e.t.c. (Goldblatt et al. 2013). bauer et al. 1988; Seyfried 1987), because Earth seawater We neglect tidal heating, because there is no evidence is Na-rich (0.5 mol/kg). In summary, waterworld ocean that tidal heating exceeds radiogenic heating for any HZ Na concentrations O(0.1-1) mol/kg are geologically rea- planet around a Sun-like star after 100 Myr. sonable. Geophysics. In order to make our calculations simpler, 6.6. Opportunities we understate the pressure needed to completely sup- press volcanism. This pressure may be closer to 3 GPa Our analysis suggests the following science opportuni- than 1 GPa (Noack et al. 2016). ties for waterworld research: We assume stagnant-lid tectonics. We suspect stag- • More detailed modeling of water-rock interaction nant lid is the default mode of planetary tectonics, in throughout waterworld evolution, using newly- part because stagnant lid worlds outnumber plate tec- available constraints (e.g., Pan et al. 2013). tonic worlds by many to one in the Solar System, and in • New measurements (via numerical and/or lab exper- part because of the difficulty of explaining plate tectonics iments) of solubilities, e.t.c., for high-pressure, low- (Korenaga 2013). Also, plate tectonics might require vol- temperature conditions relevant to waterworlds. canism (Sleep 2000), and volcanism is curtailed on water- worlds (Kite et al. 2009). If plate tectonics nevertheless • Map out rocky-planet atmospheres and atmosphere- operates on waterworlds, then relative to stagnant lid ocean equilibria in volatile-abundance space, taking mode the lithosphere might store less C. Plate tectonics into improving constraints on the galactic range of involves a thin carbonated layer that is continually sub- H/Cl, H/N, and H/C (Bergin et al. 2015). Our results ducted and decarbonated (Kelemen & Manning 2015), are sensitive to the mean value of cC /fW . whereas stagnant-lid mode slowly builds-up a potentially • Seek to confirm (or reject) hints of life proliferating thick layer of thermally stable carbonated rocks (Foley at T > 400K at Earth seafloor hydrothermal vents & Smye 2017). For a planet with plate tectonics, the (Schrenk et al. 2003). Based on the results of Fig. 10, mass of rock that interacts with the ocean can be large confirmation would underline the habitability of fW = (Sleep & Zahnle 2001). Increased rock supply makes it 0.1 worlds, regardless of whether or not HP-ice is suf- more likely that ocean chemistry would reach equilibrium ficient to prevent habitability. with minerals (Sillen 1967). If plate tectonics operated • Investigate slurry planets. Under giant-impact condi- without volcanism, then the seafloor would be composed tions, rock and water are fully miscible. We term this of mantle rocks whose aqueous alteration would produce mix a “slurry.” On planets with fW  0.01, the slurry copious H2. Copious H2 production would also occur cools quickly, and the rock settles out, because rock is if the crust extruded at the end of magma ocean crys- insoluble in cool water. Hot-slurry-layer lifetime can- tallisation was very olivine-rich (Sleep et al. 2004). This not exceed the conductive cooling timescale, might affect atmospheric composition and climate.  2 Geochemistry and climate. We use results from the 1D ρcp,f ∆Z∆zbl ∆Z τslurry . 2 ∼ 1 Gyr (8) climate model of Wordsworth & Pierrehumbert (2013) in (2.32) kcond 200km this study (e.g., Fig. 1). Recent 3D GCM work validates the trends shown in Fig. 1 (e.g. Leconte et al. 2013; Wolf where ρ is characteristic fluid density (∼1100 kg m−3), 3 −1 −1 & Toon 2015; Popp et al. 2016; Wolf et al. 2017; Kop- cp,f is fluid heat capacity (∼4 × 10 J kg K ), ∆Z parapu et al. 2017). However, the amplitudes, location, is ocean depth, ∆zbl is the thickness of the conduc- and even existence of the instabilities discussed in §4 may tive boundary layer, kcond is characteristic fluid ther- depend on the specifics of the climate model (e.g. Yang mal conductivity (1 W m−1 K−1), and we have made et al. 2016). the upper-limit assumption ∆Z = ∆zbl (Turcotte & Our treatment of both carbonate formation and CO2 Schubert 2011). On planets with large fW , a parcel solubility in cation-rich fluids in §4 is crude and this could of the dense, hot, rock+water layer is stable to rapid shift the optimum cC for habitability by a factor of ∼2. convective swapping with the overlying cool rock-poor Habitability of exoplanet waterworlds 19

aqueous fluid. Convective inhibition slows slurry cool- Atmosphere-lithosphere geochemical cycling appears ing and allows super-adiabatic temperature gradients to be necessary for Earth’s long-term habitability (Moore (persistence of hot layers at depth). Thus, an ocean et al. 2017). This has led to the idea that “[g]eologic at habitable temperature might be underlain not by a activity is crucial for a planet’s maintained surface hab- solid seafloor but instead by a hot slurry. The true life- itability because such habitability depends on the recy- time will be set by double-diffusive convection (Radko cling of atmospheric gases like CO2” (Kaltenegger 2017). 2013; Friedson & Gonzales 2017; Moll et al. 2017), tak- However, as we have shown here, such cycling is not ing account of the exotic behavior of aqueous fluids at needed for multi-Gyr habitability on rocky exoplanets P > 10 GPa (e.g., Redmer et al. 2011; Tian & Stanley with deep oceans. 2013), and could be much less than the Eqn. 8 upper Our optimistic conclusions for Sunlike stars suggest limit. It is not obvious whether such a system would similar optimism for waterworlds that orbit M-dwarfs be more habitable or less habitable than a fluid ocean (Turbet et al. 2017b). For Sunlike stars, stellar main- underlain by a solid crust. sequence evolution (timescale τMS ∼10 Gyr) ultimately destroys habitability. But for planets around M-stars, • Combine eH modeling (Schaefer et al. 2016; τMS  10 Gyr. Because volcanism-requiring (Earth-like) Wordsworth et al. 2017) with pH modeling (this study). habitability dies with volcanism after .10 Gyr, most of the habitable volume in the Universe (including the dis- • Evaluate the climate feedbacks proposed in §6.2. Ex- tant future; Loeb et al. 2016) is for cycle-independent plore the bestiary of exsolution-driven climate feed- planetary habitability. For example, the TRAPPIST-1 backs suggested by our simple model (§4, Fig. 7, Kite habitable-zone planets are small enough and old enough et al. 2011). (Burgasser & Mamajek 2017) that stagnant-lid volcan- ism should have shut down (Kite et al. 2009). Thus the 7. CONCLUSIONS mechanisms proposed in this paper offer more hope for Our model indicates that the key controls on the hab- habitability in the TRAPPIST-1 system than volcanism- itability of waterworlds around Sunlike stars are: requiring habitability. However, although Earth-radius waterworlds in the HZ of a Sun-like star can retain their 1. Initial water content. As ocean depth is increased, water, it is not clear whether or not this is the case for long-term atmosphere-lithosphere cycling of C moves Earth-radius waterworlds in the HZ of a M star (Dong from an active regime, to a subdued regime, to an ex- et al. 2017). Although in the absence of volcanism, otic regime. We studied planets with 1-8 GPa seafloor nutrient supply will ultimately become diffusively lim- pressure. Seafloor pressure on such worlds suppresses ited, habitable surface water on waterworlds can persist rock melting, and so atmosphere-lithosphere cycling is for 10 Gyr. For these worlds, habitability has no need subdued (Fig. 3) – permitting cycle-independent plan- for geodynamic processes, but only the steady light of etary habitability (Fig 6). the star. 2. Cation/C ratio. [Cation]/C values . 1 are needed (in our model) for > 1 Gyr of habitability, because a cation excess draws C into the ocean from the atmosphere, We are grateful to Bruce Fegley for continuing guidance and worlds with little atmospheric C have short life- on geochemical thermodynamics. We thank Itay Halevy, times (Figs. 14-15). Cation supply is set by leaching Nataly Ozak, Ravi Kumar Kopparapu, Jim Kasting, and from the crust. Leaching is limited by crustal thickness Christophe Cossou, for unselfish sharing of research out- and the intrusive/extrusive ratio. put in the early stages of this project. We thank Do- rian Abbot, David Archer, Paul Asimow, Fred Ciesla, 3. Initial C/H content. When C/H is large, C sinks are Jim Fuller, Eric Gaidos, Marc Hirschmann, Nade- swamped and CO2 accumulates in the atmosphere. jda Marounina, Mohit Melwani Daswani, Jack Mus- This can lead to uninhabitably high surface temper- tard, Leslie Rogers, Laura Schaefer, Hilke Schlichting, atures (Fig. 13). When C/H is small, the duration of Norm Sleep, Sarah Stewart, Cayman Unterborn, and surface liquid water is < 1 Gyr. C/H ≈ 1% by weight Steve Vance, who each provided useful advice. We thank is optimal in our model (corresponding to CO2/H2O the CHIM-XPT team: Mark Reed, Nicolas Spycher, ≈ 0.25% by weight if all C is present as CO2 and all H and Jim Palandri. We thank Nadejda Marounina for is present as H2O) (Figs. 11-15). C/H  1 is expected commenting on a draft, and Craig Manning for shar- on H2-free rocky planets with deep oceans, because C ing a preprint. We thank the organizers of the Univer- is iron-loving under giant impact conditions and so is sity of Michigan volatile origins workshop. The Center permanently sequestered into the liquid metal. for Exoplanets and Habitable Worlds is supported by the Pennsylvania State University, the Eberly College of ∼1/4 of waterworlds in our Monte Carlo simulations Science, and the Pennsylvania Space Grant Consortium. maintain habitable surface water for >1 Gyr. We as- Portions of this research were conducted with Advanced sume that initially frozen surfaces are darkened by mete- CyberInfrastructure computational resources provided oritic debris; if initially-frozen surfaces maintain a high by The Institute for CyberScience at The Pennsylvania albedo, then the proportion of long-lived worlds is cut State University (http://ics.psu.edu), including the Cy- by a factor of ∼3. This does not mean that in the real berLAMP cluster supported by NSF grant MRI-1626251. galaxy, &10% of waterworlds are habitable for >1 Gyr, The results reported herein benefitted from collabora- because many key input parameters are uncertain. Based tions and/or information exchange within NASA’s Nexus on these uncertainties, we recommend priority areas for for Exoplanet System Science (NExSS) research coordi- future research (§6.6). nation network sponsored by NASA’s Science Mission 20 Kite & Ford

Directorate. This work was supported by the U.S. tax- payer, primarily via NASA grant NNX16AB44G.

Author contributions: E.S.K. conceived, designed, and carried out the study, and wrote the paper. E.S.K. wrote the volatile tracking code (§5.2). E.B.F’s contributions were focused on the N-body integrations (§5.1).

APPENDIX A. HOW SEAFLOOR PRESSURE SUPPRESSES VOLCANISM. With reference to Fig. 3, four effects allow seafloor pres- sure to suppress volcanism on waterworlds. (1) The solidus T increases rapidly with P , so increasing P trun- cates the maximum melt fraction, eventually to zero. (2) Upwelling mantle undergoes corner flow (turns side- ways), so by reducing the distance between the base of the lithosphere and the maximum P for melting, in- creasing P reduces the proportion of upwelling mantle that melts. (3) T just below the lithosphere is regu- lated to an absolute value that is fairly insensitive to the ocean depth, due to a negative feedback between T - dependent mantle viscosity and the viscosity-dependent rate of convective mantle cooling (Korenaga & Karato 2008; Stevenson 2003). Therefore, a waterworld man- tle is colder for a given P (dashed blue lines in Fig. 3) than the mantle of a shallow-ocean planet (dashed gray lines in Fig. 3). This inhibits melting even more than is implied by comparing tracks A-B-C-D and A-B-C0-D0 on Fig. 3. (4) For the melt fractions corresponding to tracks A-B-C-D and A-B-C0-D0, melt fraction increases 1.5 1.5 as (T −Tsol) /(Tliq −Tsol) (Katz et al. 2003). There- fore, the melt fraction at C greatly exceeds the melt frac- tion at C0. Figure 18. Seafloor basalt alteration for (upper panel) low [C] We use the anhydrous batch-melting solidus of Katz concentrations and (lower panel) high [C] concentrations. W/R is water/rock ratio characterizing the reaction. The thick colored et al. (2003). Plausible waterworld-upper-mantle wa- lines correspond to concentrations in the fluid. Thin lines of the ter contents are similar to those of the Earth’s man- same color correspond to total mol/kg of rock added to the fluid. tle beneath mid-ocean ridges. These rocks have Where the total mol/kg of rock added from the fluid deviates from 0.005-0.02 wt% water, and this water might depress the the concentrations in the fluid, secondary minerals have formed. solidus by 50K. The thin black line corresponds to -log10(pH).

B. HIGH-PRESSURE ICE PHASES. C. POST-MAGMA-OCEAN TIMESCALES.

HP-ice appears wherever the H2O adiabat intersects the The characteristic T at which water first reacts with the pure-H2O freezing curve. (We neglect salt effects; Vance crust depends on the relative timescales of fluid-envelope & Brown 2013.) To build the fluid-water adiabat, we cooling (τc) and of crust production (τx). If τc < τx, used crustal rocks will first react with hot (supercritical) water (Cannon et al. 2017). If τc > τx, crustal rocks will react ∂T α(T, p)T with water on an adiabat connected to an sea-surface ad = (B1) ∂P ρ(T i, pi)Cp that is in equilibrium with insolation (Fig. 1.1). The timescale for cooling τc is just where ρ is obtained from Table 1 of Wiryana et al. (1998) and Table 2 of Abramson & Brown (2004), and MplfW cp,f (2000K − 800K) τc = 2 (C1) α is obtained by differencing these tables. We asumed 4πR (FKI − 0.25L∗(1 − α)) −1 −1 pl Cp ∼ 3800 J kg K . The approximation of Cp as constant is justified by Myint et al. (2017). We checked where cp,f is the fluid heat capacity and FKI is the 5 6 the resulting adiabat against Fig. 1b of Myint et al. Komabayashii-Ingersoll limit. This gives τc = 10 -10 (2017). We used extrapolation above 6 GPa , and for yr for the fluid envelopes considered in this paper. The ◦ T < 80 C. Clathrates were neglected. Pure-H2O freez- timescale for crust formation τx is – in general – not well ing curves for ices VI and VII are from IAPWS (2011). constrained. τx for stagnant lid mode on waterworlds is We note that Alibert (2014) did not include the adia- the timescale for production of the initial post-magma- batic heating of liquid water, which is a key effect for ocean crust. During this process, any given melt parcel deep oceans (Fig. 1.1). is extracted to the crust in  1 Myr, but melting might Habitability of exoplanet waterworlds 21 continue for 108 yr (Rubin et al. 2005; Mezger et al. 2013; Abe, Y., Abe-Ouchi, A., Sleep, N. H., & Zahnle, K. J. 2011, Moynier et al. 2010; Turner et al. 2001; Solomatov 2015). Astrobiology, 11, 443 Abramson, E. H., & Brown, J. M. 2004, Also relevant is the timescale for C delivery (τdyn). τdyn 8 Geochim. Cosmochim. Acta, 68, 1827 could be >10 yr for the Earth, and this may be typical Akanuma, S., Nakajima, Y., Yokobori, S.-i., et al. 2013, PNAS, of planets around Sunlike stars (L¨ohneet al. 2008). Be- 110, 11067 cause τx and τdyn are both plausibly much longer than τc, Albar`ede,F. 2009, Nature, 461, 1227 but not plausibly shorter than τ , it is more likely than Alibert, Y. 2014, A&A, 561, A41 c Alt, J. C. 1995, AGU Geophys. Monograph Series, 91, 85 not that the characteristic temperature for basalt rock Archer, D., Eby, M., Brovkin, V., et al. 2009, Ann. Rev. Earth to first encounter C-rich fluid is <650 K. Water-rock re- Planet. Sci., 37, 117 actions are well advanced within 103-105 yr (Hannington Atreya, S. K., Lorenz, R. D., & Waite, J. H. 2010, in Brown, R., et al. 2005), so we expect that even τx  1 Myr would Lebreton, J. L., & Waite, H., (eds.), Titan from allow sufficient time for alteration. Cassini-Huygens, 177 Bahcall, J. N., Pinsonneault, M. H., & Basu, S. 2001, ApJ, 555, 990 D. MODELING OF SEAWATER-BASALT INTERACTION. Bains, W., Xiao, Y., & Yu, C., 2015, Life, 5(2), 1054-1100 To explore seawater-basalt reactions on waterworlds, Ballard, S., Chaplin, W. J., Charbonneau, D., et al. 2014, ApJ, 790, 12 we used CHIM-XPT v.2.4.6 (Reed 1998). We con- Batygin, K., & Laughlin, G. 2015, PNAS, 112, 4214 sider T = {298K, 473K, 573K}. Although tempera- Beerling, D. 2008, The Emerald Planet, Oxford University Press ture changes over geologic time, we assume no “reset- Bergin, E. A., Blake, G. A., Ciesla, F., et al., 2015, PNAS, 112, ting” of seafloor alteration. “Resetting” appears to be 8965 minor for Earth seafloor alteration: seafloor basalt car- Bl¨attler,C. L., Kump, L. R., Fischer, W. W., et al. 2017, Nature Geoscience, 10, 41 bonitization occurs near the time of eruption (Coogan Blount, Z. D., Borland, C. Z., & Lenski, R. E. 2008, PNAS, 105, et al. 2016). Due to the scarcity of experiments on H2O 7899 properties for >0.5 GPa, P is fixed at 0.5 GPa. This Bollengier, O., Choukroun, M., Grasset, O., et al. 2013, low P will cause us to understate solubilities. Improved Geochim. Cosmochim. Acta, 119, 322 ab initio calculations of the properties of H O (Pan et Broecker, W. S., & Sanyal, A. 1998, Global Biogeochemical 2 Cycles, 12, 403 al. 2013) will allow future improvements in waterworld Bullock, M. A., & Moore, J. M. 2007, Geophys. Res. Lett., 34, seafloor modeling. We assume thermodynamic equilib- L19201 rium, with solid solutions suppressed. Basalt compo- Burgasser, A. J., & Mamajek, E. E. 2017, ApJ, 845, 110 sition is from Gale et al. (2013). A key parameter in Burger, C., Maindl, T. I., & Sch¨afer,C. M. 2017, CHIM-XPT is the water/rock mass ratio, W/R. The arXiv:1710.03669 Burke, C. J., Christiansen, J. L., Mullally, F., et al. 2015, ApJ, use of W/R by geochemists is easy to misinterpret in 809, 8 an exoplanet context. W/R corresponds to the effective Caldeira, K., & Kasting, J. F. 1992, Nature, 359, 226 W/R of the reactions. Reactions typically occur in zones Cannon, K. M., Parman, S. W., & Mustard, J. F. 2017, Nature that are temporarily isolated from the ocean, e.g. pore Carroll, J. J., Slupsky, J. D., & Mather, A. E. 1991, Journal of spaces and fractures. As a result, the reactions tend Physical and Chemical Reference Data, 20, 1201 Carter, B. 1983, RSPTA, 310, 347 to be much more rock-influenced than the ocean is a Catling, D. C., & Kasting, J. F. 2017, Atmospheric Evolution on whole. Therefore, W/R is almost always much less than Inhabited and Lifeless Worlds, Cambridge University Press the (mass of the ocean)/(total mass of altered rock). For Ciesla, F. J., Mulders, G. D., Pascucci, I., & Apai, D. 2015, ApJ, example, Earth hydrothermal vents have W/R of 1-100. 804, 9 Varying W/R is a proxy for the extent of alteration of Clay, P. L., Burgess, R., Busemann, H., et al. 2017, Nature, 551, 614 sub-oceanic crust by water. W/R is more important than Coogan, L. A., Parrish, R. R., & Roberts, N. M. W. 2016, T for setting vent fluid compositions. Geology, 44, 147 Fig. 18 shows typical results. For W/R < 100 char- COEL (Committee on the Origins and Evolution of Life), 2007, acteristic of hydrothermal reactions on Earth, daughter The Limits of Organic Life in Planetary Systems, National + Research Council, ISBN: 0-309-66906-5, 116 pp. fluids are dominated by Na and by dissolved-Si species. Cossou, C., Raymond, S. N., Hersant, F., et al., 2014, A&A, 569, However, dissolved Si forms solid particles upon mixing A56 into the colder, lower-pH upper levels of the ocean (as Dahl, T. W., & Stevenson, D. J. 2010, E&PSL , 295, 177 is inferred to occur on Enceladus: Hsu et al. 2015), and Dasgupta, R. 2013, Rev. Mineral. Geochem., 75, 183 we do not consider dissolved Si when calculating the sea- Dolejˇs,D., & Manning, C. E. 2010, Geofluids, 10, 20–40 Dong, C., Huang, Z., Lingam, M., et al. 2017, ApJ, 847, L4 surface chemistry. Carbonate formation buffers C fluid Dorn, C., Khan, A., Heng, K., et al. 2015, A&A, 577, A83 content to ∼0.1 mol/kg for W/R . 30. Secondary min- Dorn, C., Bower, D. J., & Rozel, A. B. 2017, arXiv:1710.05605 eral formation scrubs Fe2+, Mg2+, and Al3+ from the Dressing, C. D., & Charbonneau, D. 2015, ApJ, 807, 45 fluid. Outlet fluid composition leveled off at [Na] ≈ 0.05- Duan, Z., & Sun, R. 2003, Chemical Geology, 193, 257 Dwardakas, V.V., Dauphas, N., Meyer, B., Boyajian, P., & 0.12 mol/kg for W/R ≈ 10 due to Na-mineral formation, Bojazi, M. 2003, ApJ, 851:147 but for reasons discussed in §6.5, we consider higher [Na] Edmond, J. M., & Huh, Y. 2003, E&PSL , 216, 125 in our models. Edwards, C. S., & Ehlmann, B. L. 2015, Geology, 43, 863 Fegley, B., Jr., Jacobson, N. S., Williams, K. B., et al. 2016, ApJ, 824, 103 REFERENCES Fisher, A. T. 1998, Rev. Geophys., 36, 143 Foley, B. J. 2015, ApJ, 812, 36 Abbot, D. S., Cowan, N. B., & Ciesla, F. J. 2012, ApJ, 756, 178 Foley, B. J., & Smye, A. J. 2017, arXiv:1712.03614 Abbot, D. S., Voigt, A., & Koll, D. 2011, JGR-A, 116, D18103 Ford, E. B., & Rasio, F. A. 2008, ApJ, 686, 621-636 Abe, Y., Ohtani, E., Okuchi, T., Righter, K., & Drake, M. 2000, Friedson, A. J., & Gonzales, E. J. 2017, Icarus, 297, 160 p.413- in Canup, R.M., & K. Righter (Eds.) Origin of the Fu, R., O’Connell, R. J., & Sasselov, D. D. 2010, ApJ, 708, 1326 Earth and Moon, University of Arizona Press. 22 Kite & Ford

Gaidos, E., Deschenes, B., Dundon, L., et al. 2005, Astrobiology, Knoll, A. H., Canfield, D. E., & Konhauser, K. O. (editors) 2012, 5, 100 Fundamentals of Geobiology, Wiley-Blackwell, 2012, Gaidos, E., Krot, A. N., Williams, J. P., & Raymond, S. N. 2009, Kodama, T., Genda, H., Abe, Y., & Zahnle, K. J. 2015, ApJ, 812, ApJ, 696, 1854 165 Galvez, M. E., Connolly, J. A. D., & Manning, C. E. 2016, Komacek, T. D., & Abbot, D. S. 2016, ApJ, 832, 54 Nature, 539, 420 Kopp, R. E., Kirschvink, J. L., Hilburn, I. A., & Nash, C. Z. Galbraith, E. D., & Eggleston, S. 2017, Nature Geoscience, 10, 2005, PNAS, 102, 11131 295 Korenaga, J., & Karato, S.-I. 2008, JGR (Solid Earth), 113, Gale, A., Dalton, C. A., Langmuir, C. H., Su, Y., & Schilling, B02403 J.-G. 2013, Geochemistry, Geophysics, Geosystems, 14, 489 Korenaga, J. 2013, Ann. Rev. Earth and Planet. Sci., 41, 117 Galy, V., Peucker-Ehrenbrink, B., & Eglinton, T. 2015, Nature, Korenaga, J., Planavsky, N. J., & Evans, D. A. D. 2017, Phil. 521, 204 Trans. Roy. Soc. A, 375, 20150393 Genda, H., & Abe, Y. 2005, Nature, 433, 842 Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, Gillon, M., Triaud, A. H. M. J., Demory, B.-O., et al. 2017, 765, 131 Nature, 542, 456 Kopparapu, R. K., Ramirez, R. M., SchottelKotte, J., et al. 2014, Glein, C. R., Baross, J. A., & Waite, J. H. 2015, ApJ, 787, L29 Geochim. Cosmochim. Acta, 162, 202 Kopparapu, R. k., Wolf, E. T., Arney, G., et al. 2017, ApJ, 845, 5 Golabek, G. J., Emsenhuber, A., Jutzi, M., Asphaug, E. I., & Krissansen-Totton, J., & Catling, D. C. 2017, Nature Gerya, T. V. 2017, arXiv:1710.03250 Communications, 8, 15423 Goldblatt, C., Robinson, T. D., Zahnle, K. J., & Crisp, D. 2013, Kuchner, M. J. 2003, ApJ, 596, L105 Nature Geoscience, 6, 661 Kuramoto, K., & Matsui, T. 1996, J. Geophys. Res., 101, 14909 Goldblatt, C. 2015, Astrobiology, 15, 362 Lacki, B. C. 2016, arXiv:1609.05931 Gounelle, M. 2015, A&A, 582, A26 Lammer, H., Odert, P., Leitzinger, M., et al. 2009, A&A, 506, 399 Grotzinger, J. P., & Kasting, J. F. 1993, J. Geol., 101, 235 Leconte, J., Forget, F., Charnay, B., Wordsworth, R., & Pottier, Gysi, A. P., & Stef´ansson,A. 2012, Chemical Geology, 306, 10 A. 2013, Nature, 504, 268 Halevy, I., & Schrag, D. P. 2009, Geophys. Res. Lett., 36, L23201 L´eger,A., Selsis, F., Sotin, C., et al. 2004, Icarus, 169, 499 Halevy, I., & Bachan, A. 2017, Science, 355, 1069 Levi, A., Sasselov, D., & Podolak, M. 2014, ApJ, 792, 125 Hayes, J.M., & Waldbauer, J.R. 2006, Phil. Trans. Roy. Soc. B, Levi, A., Sasselov, D., & Podolak, M. 2017, ApJ, 838, 24 361(1470), 931-50 Levison, H. F., Kretke, K. A., Walsh, K. J., et al. 2015, PNAS, Hamano, K., Abe, Y., & Genda, H. 2013, Nature, 497, 607 112, 14180 Hand, K. P., & Chyba, C. F. 2007, Icarus, 189, 424 Lichtenberg, T., Parker, R. J., & Meyer, M. R. 2016, MNRAS, Hannington, M. D., de Ronde, C. E. J., & Peterson, S. 2005, 462, 3979 Economic Geology, 100th Anniversary Volume, 111-141 Loeb, A., Batista, R. A., & Sloan, D. 2016, JCAP, 8, 040 Hartmann, L., Ciesla, F., Gressel, O., & Alexander, R. 2017, L¨ohne,T., Krivov, A. V., & Rodmann, J. 2008, ApJ, 673, arXiv:1708.03360 1123-1137 Herzberg, C., Condie, K., & Korenaga, J. 2010, E&PSL , 292, 79 Luger, R., Sestovic, M., Kruse, E., et al. 2017, Nature Astronomy, Hoffman, P.F., Abbot, D.S., Ashkenazy, Y., et al. 2017, Science 1, 0129 Advances, 3, DOI: 10.1126/sciadv.1600983. Lupu, R. E., Zahnle, K., Marley, M. S., et al. 2014, ApJ, 784, 27 Holland, H. D. 1984, The chemical evolution of the atmosphere MacLeod, G., McKeown, C., Hall, A. J., & Russell, M. J. 1994, and oceans. Princeton University Press, Princeton, N.J., USA. Origins of Life and Evolution of the Biosphere, 24, 19 Hirschmann, M. M. 2006, Ann. Rev. Earth Planet. Sci., 34, 629 Maher, K., & Chamberlain, C. P. 2014, Science, 343, 1502 Hirschmann, M. M. 2012, E&PSL, 341, 48 Maindl, T. I., Sch¨afer,C. M., Haghighipour, N., Burger, C., & Hirschmann, M. M. 2016, American Mineralogist, 101, 540 Dvorak, R. 2017, Proc. of the 1st Greek-Austrian Workshop on Hsu, H.-W., Postberg, F., Sekine, Y., et al. 2015, Nature, 519, 207 Extrasolar Planetary Systems, 137 Hu, Y., Yang, J., Ding, F., et al., 2011, Clim. Past, 7, 17 Marcus, R. A., Sasselov, D., Stewart, S. T., & Hernquist, L. The International Association for the Properties of Water and 2010a, ApJ, 719, L45 Steam, 2011, Revised Release on the Pressure along the Marcus, R. A., Sasselov, D., Hernquist, L., & Stewart, S. T. Melting and Sublimation Curves of Ordinary Water Substance, 2010b, ApJ, 712, L73 IAPWS R14-08(2011). Marty, B. 2012, Earth and Planetary Science Letters, 313, 56 Inamdar, N. K., & Schlichting, H. E. 2016, ApJ, 817, L13 McKinnon, W. B., & Zolensky, M. E. 2003, Astrobiology, 3, 879 Jacobsen, S. B., Ranen, M. C., Petaev, M. I., et al. 2008, RSPTA, Melosh, J. 2011, Planetary Surface Processes, Cambridge 366, 4129 University Press. Jacobson, S. A., Rubie, D. C., Hernlund, J., Morbidelli, A., & Mezger, K., Debaille, V., & Kleine, T. 2013, Space Sci. Rev., 174, Nakajima, M. 2017, E&PSL , 474, 375 27 Jura, M., Xu, S., & Young, E. D. 2013, ApJ, 775, L41 Marounina, N., & Rogers, L.A., 2017a, poster at The Origins of Izidoro, A., Ogihara, M., Raymond, S. N., et al. 2017, MNRAS, Volatiles in Habitable Planets, workshop held in Ann Arbor, MI 470, 1750 Marounina, N., Grasset, O., Tobie, G., & Carpy, S. 2017b, Kah, L. C., & Riding, R. 2007, Geology, 35, 799 arXiv:1711.09128 Kalousov´a,K., Sotin, C., Choblet, G., Tobie, G., & Grasset, O. Moll, R., Garaud, P., Mankovich, C., & Fortney, J. J. 2017, ApJ, 2018, Icarus, 299, 133 849, 24 Kaltenegger, L. 2017, ARA&A, 55, 433 Moore, G., Vennemann, T., & Carmichael, I. S. E. 1998, Kasting, J. F. 1988, Icarus, 74, 472 American Mineralogist, 83, 36 Kasting, J. F., Whitmire, D. P., & Reynolds, R. T. 1993, Icarus, Moore, W. B., Lenardic, A., Jellinek, A. M., et al. 2017, Nature 101, 108 Astronomy, 1, 0043 Kasting, J. F., & Catling, D. 2003, ARA&A, 41, 429 Morbidelli, A., Bitsch, B., Crida, A., et al. 2016, Icarus, 267, 368 Kasting, J. F., Kopparapu, R., Ramirez, R. M., & Harman, C. E. Moynier, F., Yin, Q.-Z., Irisawa, K., et al. 2010, PNAS, 107, 2014, PNAS, 111, 12641 10810 Katz, R. F., Spiegelman, M., & Langmuir, C. H. 2003, Mulders, G. D., Ciesla, F. J., Min, M., & Pascucci, I. 2015, ApJ, Geochemistry, Geophysics, Geosystems, 4, 1073 807, 9 Kelemen, P. B., & Hirth, G. 2012, E&PSL, 345, 81 Myint, P. C., Benedict, L. X., & Belof, J. L. 2017, Kelemen, P. B., & Manning, C. E. 2015, PNAS, 112, E3997 J. Chem. Phys., 147, 084505 Kempe, S., & Degens, E. T. 1985, Chemical Geology, 53, 95 Nakamura, K., & Kato, Y. 2004, Geochim. Cosmochim. Acta, 68, Kite, E. S., Manga, M., & Gaidos, E. 2009, ApJ, 700, 1732 4595 Kite, E. S., Gaidos, E., & Manga, M. 2011, ApJ, 743, 41 Neveu, M., Desch, S. J., & Castillo-Rogez, J. C. 2015, JGR Kitzmann, D., Alibert, Y., Godolt, M., et al. 2015, MNRAS, 452, (Planets), 120, 123 3752 Nitschke, W., McGlynn, S. E., Milner-White, E. J., & Russell, M. J. 2013, Biochim Biophys Acta 1827(8-9):871-81 Habitability of exoplanet waterworlds 23

Noack, L., H¨oning,D., Rivoldini, A., et al. 2016, Icarus, 277, 215 Stevenson, D. J. 2003, Comptes Rendus Geoscience, 335, 99 O’Brien, D. P., Walsh, K. J., Morbidelli, A., Raymond, S. N., & Stewart, S. T., Lock, S. J., & Mukhopadhyay, S. 2014, Lunar and Mandell, A. M. 2014, Icarus, 239, 74 Planetary Science Conference, 45, 2869 Odert, P., Lammer, H., Erkaev, N. V., et al. 2017, Staudigel, H., & Hart, S. R. 1983, Geochim. Cosmochim. Acta, arXiv:1706.06988 47, 337 Ogihara, M., Morbidelli, A., & Guillot, T. 2015, A&A, 578, A36 Stolper, D. A., Bender, M. L., Dreyfus, G. B., Yan, Y., & Ohtani, E. 2005, Elements 1(1), 25-30. Higgins, J. A. 2016, Science, 353, 1427 O’Rourke, J. G., & Korenaga, J. 2012, Icarus, 221, 1043 Sun, Z., Ji, J., Wang, S., & Jin, S. 2017, MNRAS, 467, 619 Owen, J. E., & Alvarez, M. A. 2016, ApJ, 816, 34 Takai, K., Nakamura, K., Toki, T., et al. 2008, PNAS, 105, 10949 Pan, D., Spanu, L., Harrison, B., Sverjensky, D. A., & Galli, G. Tart`ese,R., Chaussidon, M., Gurenko, A., et al. 2017, 2013, PNAS, 110, 6646 Geochemical Perspective Letters, 3, n1. Piso, A.-M. A., Oberg,¨ K. I., Birnstiel, T., & Murray-Clay, R. A. Taylor, S. R., & McLennan, S. 2009, Planetary Crusts. 2015, ApJ, 815, 109 Cambridge University Press Plesa, A.-C., Tosi, N., & Breuer, D. 2014, E&PSL, 403, 225 Thomas, P. C., Parker, J. W., McFadden, L. A., et al. 2005, Popp, M., Schmidt, H., & Marotzke, J. 2016, Nature Nature, 437, 224 Communications, 7, 10627 Tian, F., & Ida, S. 2015, Nature Geoscience, 8, 177 Radko, T. 2013, Double Diffusive Convection, Cambridge Tian, B.Y., & Stanley, S. 2015, ApJ, 768:156 University Press. Tikoo, S. M., & Elkins-Tanton, L. T. 2017, RSPTA, 375, Ramirez, R. M., & Kaltenegger, L. 2017, ApJ, 837, L4 20150394 Raymond, S. N., Quinn, T., & Lunine, J. I. 2004, Icarus, 168, 1 Tosi, N., Godolt, M., Stracke, B., et al. 2017, A&A, 605, A71 Raymond, S. N., Quinn, T., & Lunine, J. I. 2007, Astrobiology, 7, Turbet, M., Forget, F., Leconte, J., Charnay, B., & Tobie, G. 66 2017, E&PSL , 476, 11 Redfern, S. A. T., Salje, E., & Navrotsky, A. 1989, Contributions Turbet, M., Bolmont, E., Leconte, J., et al. 2017, to Mineralogy and Petrology, 101, 479 arXiv:1707.06927 Redmer, R., Mattsson, T. R., Nettelmann, N., & French, M. Turcotte, D. L., & Schubert, G. 2011, Geodynamics - 3rd Edition, 2011, Icarus, 211, 798 Cambridge University Press. Reed, M. H. 1998, pp. 109–124 in Techniques in Hydrothermal Turner, S., Evans, P., & Hawkesworth, C. 2001, Science, 292, 1363 Ore Deposits Geology, edited by J. P. Richards & P. B. Larson, Unterborn, C. T., Desch, S. J., Hinkel, N., & Lorenzo, A. 2017, Society of Economic Geologists, Littleton, CO. arXiv:1706.02689 Rein, H. & Liu, S.-F. 2012, ˚a,537, 128 Valencia, D., Sasselov, D. D., & O’Connell, R. J. 2007, ApJ, 665, Rein, H. & Spiegel, D.S. 2015, MNRAS, 446, 1424 1413 Ridgwell, A., & Zeebe, R. E. 2005, E&PSL , 234, 299 Vance, S., Harnmeijer, J., Kimura, J., et al. 2007, Astrobiology, 7, Rogers, L. A., & Seager, S. 2010, ApJ, 712, 974 987 Rosenbauer, R. J., Bischoff, J. L., & Zierenberg, R. A. 1988, Vance, S., & Brown, J. M. 2013, Geochim. Cosmochim. Acta, Journal of Geology, 96, 237 110, 176 Rosenberg, N. D., Browning, L., & Bourcier, W. L. 2001, M&PS, Vance, S. D., Hand, K. P., & Pappalardo, R. T. 2016, 36, 239 Geophys. Res. Lett., 43, 4871 Rothman, D. H., & Forney, D. C. 2007, Science, 316, 1325 Vermeij, G. J. 2006, PNAS, 103, 1804 Rubie, D. C., Jacobson, S. A., Morbidelli, A., et al. 2015, Icarus, Walker, J. C. G., Hays, P. B., & Kasting, J. F. 1981, 248, 89 J. Geophys. Res., 86, 9776 Rubin, K. H., van der Zander, I., Smith, M. C., & Bergmanis, White, A. F., & Brantley, S. L. 1995, Rev. Mineral. Geochem., 31. E. C. 2005, Nature, 437, 534 White, S. M., Crisp, J. A., & Spera, F. J. 2006, Geochemistry, Russell, M. J., & Arndt, N. T. 2005, Biogeosciences, 2, 97 Geophysics, Geosystems, 7, Q03010 Russell, J. F., Hall, A. J., & Martin, W. 2010, Geobiology, 8, 355 White, W.M., & Klein, E.M. 2014, Composition of the oceanic Sato, T., Okuzumi, S., & Ida, S. 2016, A&A, 589, A15 crust, chapter 4.13 in Treatise on Geochemistry (2nd Edition), Schaefer, L., Wordsworth, R. D., Berta-Thompson, Z., & H.D. Holland & K.K. Turekian, Eds, Elsevier, Oxford. Sasselov, D. 2016, ApJ, 829, 63 Williams, D. A., & Greeley, R. 1994, Icarus, 110, 196 Schlichting, H. E., Sari, R., & Yalinewich, A. 2015, Icarus, 247, 81 Wiryana, S., Slutsky, L. J., & Brown, J. M. 1998, E&PSL, 163, Schoonen, M., & Smirnov, A. 2016, Elements, 12, 395-400 123 Schrenk, M. O., Kelley, D. S., Delaney, J. R., & Baross, J.A. Wolf, E. T., & Toon, O. B. 2015, JGR (Atmospheres), 120, 5775 2003, Applied & Environmental Microbiology 69, 3580-3592. Wolf, E. T., Shields, A. L., Kopparapu, R. K., Haqq-Misra, J., & Segura, T. L., McKay, C. P., & Toon, O. B. 2012, Icarus, 220, 144 Toon, O. B. 2017, ApJ, 837, 107 Selsis, F., Chazelas, B., Bord´e,P., et al. 2007, Icarus, 191, 453 Wordsworth, R. 2012, Icarus, 219, 267 Seyfried, W. E., Jr. 1987, Ann. Rev. Earth Planet. Sci., 15, 317 Wordsworth, R. D., & Pierrehumbert, R. T. 2013, ApJ, 778, 154 Sharma, A., Scott, J. H., Cody, G. D., et al. 2002, Science, 295, Wordsworth, R. D. 2016, Ann. Rev. Earth Planet. Sci., 44, 381 1514 Wordsworth, R., Schaefer, L., & Fischer, R. 2017, Sharp, Z. D., & Draper, D. S. 2013, E&PSL , 369, 71 arXiv:1710.00345 Shibuya, T., Tahata, M., Kitajima, K., et al. 2012, E&PSL , 321, Yang, J., Leconte, J., Wolf, E. T., et al. 2016, ApJ, 826, 222 64 Yang, J., Ding, F., Ramirez, R. M., et al. 2017, Nature Shibuya, T., Yoshizaki, M., Masaki, Y., et al. 2013, Chemical Geoscience, 10, 556 Geology, 359, 1 Young, E. D. 2001, RSPTA, 359, 2095 Silburt, A., Gaidos, E., & Wu, Y. 2015, ApJ, 799, 180 Zahnle, K. J., Kasting, J. F., & Pollack, J. B. 1988, Icarus, 74, 62 Sillen, L. G. 1967, Science, 156, 1189 Zahnle, K., Arndt, N., Cockell, C., et al. 2007, Space Sci. Rev., Sleep, N. H. 2000, J. Geophys. Res., 105, 17563 129, 35 Sleep, N. H. 2000, Geologic processes and habitability, in Zahnle, K. J., Lupu, R., Dobrovolskis, A., & Sleep, N. H. 2015, Proceedings of the Pale Blue Dot 2 Workshop, May 19-21, E&PSL, 427, 1999, Ames Research Center Zahnle, K. J., & Catling, D. C. 2017, ApJ, 843, 122 Sleep, N. H., Zahnle, K., & Neuhoff, P. S. 2001, PNAS, 98, 3666 Zeebe, R. E. 2012, Ann. Rev. Earth Planet. Sci., 40, 141 Sleep, N. H., & Zahnle, K. 2001, J. Geophys. Res., 106, 1373 Zeebe, R. E., & Wolf-Gladrow, D. 2001, CO2 in Seawater: Sleep, N. H., Meibom, A., Fridriksson, T., Coleman, R. G., & Equilibrium, Kinetics, Isotopes, vol. 65, Elsevier Oceanography Bird, D. K. 2004, PNAS, 101, 12818 Series Sleep, N. H., Zahnle, K. J., & Lupu, R. E. 2014, RSPTA, 372, Zeebe, R. E., & Caldeira, K. 2008, Nature Geoscience, 1, 312 20130172 Zhou, J.-L., & Lin, D. N. C. 2007, ApJ, 666, 447 Solomatov, V. S. 2015, in Treatise on Geophysics (2nd. Edn). Zolotov, M. Y. 2007, Geophys. Res. Lett., 34, L23203 Stamenkovi´c,V., & Seager, S. 2016, ApJ, 825, 78 Zolotov, M. Y., & Kargel, J. S. 2009, p.431–457 in Europa, edited Stevenson, D. J. 1999, Nature, 400, 32 by R.T. Pappalardo, W.B. McKinnon, K.K. Khurana; University of Arizona Press.